Characterization of a metallic powder bed by colorimetry

Abstract

The invention relates to a method (20) for determining the oxygen concentration (Cox) of a powder (MP) of a metallic material (Mat) having a powder bed shape (PB), said method comprising the steps of: A) Acquiring an image (Im) of at least a portion of said powder bed, said image comprising a set of pixels (P), each pixel having a color coded according to a colorimetric coding comprising three quantities (G1, G2, G3); B) Determining the oxygen concentration of the powder from values ​​of said three quantities associated with pixels of the image, using a predefined calibration function (CFMat), a function of said material, and relating said oxygen concentration to said three quantities. Figure 2

Description

Title of the invention: Characterization of a bed of metallic powder by colorimetry FIELD OF INVENTION

[0001] The present invention relates to the field of metal powder characterization, more particularly to the determination of the oxygen concentration of the powder, the latter having a flat surface (powder bed). This characterization is typically of interest when implementing a metal additive manufacturing process for a part. STATE OF THE ART

[0002] When using a metal powder, it is susceptible to oxidation, which can disrupt its use. This problem arises, for example, during the implementation of a metallic additive manufacturing process by selective powder bed fusion (or MAM-PBF for Metallic Additive Manufacturing Powder Bed Fusion). The fields in which this process is typically used are aerospace, biomedical, automotive, and nuclear.

[0003] Metal additive manufacturing consists of producing parts by successively adding (metallic) material layer by layer, from a 3D digital file. A powder bed (PB) is defined as a controlled thickness of powder (MP) with a flat surface. The metallic material (Ma) constituting the powder is chosen, for example, from: stainless steels, titanium-based alloys, aluminum-based alloys, and nickel-based alloys.

[0004] The manufacturing process involves spreading thin layers of powder (typically between 10 and 100 µm thick) one on top of the other, with a selective consolidation step between each layer deposition. Selective consolidation is achieved, for example, with one or more laser beams, an electron beam, laser sintering, or binder spraying. Consolidation refers to the process of making the material rigid by bonding the powder particles together.

[0005] An example of a system 15 implementing MAM-PBF fabrication of a part Pa with an LB laser beam (referred to as L-PBF) is illustrated [Fig. 1]. It comprises a PC reservoir of metal powder MP, a manufacturing platform MPL, and a powder recovery tray PCT. A powder spreading device PSD pushes the powder from the reservoir towards the platform MPL in a spreading direction SD so as to deposit a controlled powder thickness (typically on the order of 10 to 100 µm) referred to as the powder bed PB. The SD direction is along the X-axis, and the plane of the powder bed is the XY plane, with Z being the vertical direction. The excess powder during The powder bed is poured into the PCT recovery tank. The PSD spreading device is, for example, a scraper, roller, or brush. A vertical displacement device MD for the manufacturing substrate Sub is positioned in the Tk tank used for fabrication. This device lowers the substrate and the part being formed Pa along the Z axis as the layers are deposited and consolidated. At the beginning of the fabrication process, the Sub substrate receives the powder bed, and during the process, the part is attached to this substrate. The consolidation device CD includes a laser L and an array of LSD mirrors controlled by the UTO processing unit, which deflects the laser beam LB according to the data in the 3D file.

[0006] Thus, the MAM-PBF process comprises, for each layer, a cycle including a powder spreading step involving the removal of excess powder (and the return of the PSD device to its initial position), and a consolidation step. Then, the substrate is lowered and a new cycle produces the next layer, until the final layer.

[0007] Once manufacturing is complete, the substrate to which the manufactured part(s) are attached is removed from the machine. The powder filling the Tk reservoir and the PCT bin is recovered and, if necessary, sieved, characterized, and, where appropriate, mixed with fresh powder for reuse in future manufacturing. Indeed, only a very small portion of the powder is consolidated into a part during manufacturing, and the unfused powder can be recovered and reused for future production runs. However, some particles are degraded due to interactions between the laser and the material, the high temperatures experienced, and the imperfectly controlled atmosphere in the manufacturing chamber. Thus, reusing the powder results in a degradation of the attributes of some particles, which are randomly distributed within the powder.The degradation of powder quality leads to a decrease in the properties of the manufactured parts. An increase in oxygen content in recycled powder has been observed for many materials in the literature. Some powder particles become surrounded by a more or less homogeneous oxide layer, the chemical nature of which depends on the material. For example, the publication by T. Delacroix et al., "Influence of powder recycling on 316L stainless steel feedstocks and printed parts in laser powder bed fusion" (Addit. Mauf., vol. 25, pp. 84–103, 2019), describes this phenomenon for stainless steel powders. Thus, a good indicator of powder quality degradation, typically when used for MAM-PBF manufacturing, is its oxidation.

[0008] Currently, powder samples are collected between manufacturing processes and characterized ex-situ on small quantities in order to assess their quality. These small quantities of characterized material are not necessarily representative of the entire powder.

[0009] In some cases, to avoid taking any risks, the unfused powder recovered after manufacturing is not checked but directly discarded in order to use only new powder and ensure the quality of the manufactured parts, which leads to an increase in the cost of the part.

[0010] One object of the present invention is to overcome the aforementioned drawbacks by providing a method for determining the oxygen concentration of a powder of a metallic material in the form of a powder bed, and an associated device, that is rapid, inexpensive, and can be implemented directly online during the execution of a metal powder bed additive manufacturing process. DESCRIPTION OF THE INVENTION

[0011] The present invention relates to a method for determining the oxygen concentration of a powder of a metallic material having a powder bed shape, the method comprising the steps of: - A) To produce an image of at least a portion of said powder bed, said image comprising a set of pixels, each pixel having a color coded according to a colorimetric coding comprising three quantities, and - B) Determine the oxygen concentration of the powder from values ​​of said three quantities associated with pixels of the image, using a predefined calibration function, a function of said material, and relating said oxygen concentration and said three quantities.

[0012] According to one embodiment, the colorimetric coding is the RGB system, the three quantities being named R, G and B.

[0013] According to one embodiment, the calibration function is a first or second degree polynomial with three variables corresponding to the three quantities, and whose coefficients are a function of the material.

[0014] According to one variant, step A is carried out in one go with a camera.

[0015] According to another variant, step A is carried out by scanning with a flatbed scanner.

[0016] According to a first method, step B comprises the substeps consisting of: - Bl) Determine, for a plurality of pixels in the image, an oxygen concentration called the pixel concentration, using the calibration function, and - B2) Determine the oxygen concentration from an average of the aforementioned pixel concentrations.

[0017] According to a second method, step B comprises the substeps consisting of: - B' 1) Determine an average value for each quantity from the values ​​of said quantities associated with a plurality of pixels in the image, and - B'2) Determine the oxygen concentration from the aforementioned values averages of the three quantities via the calibration function.

[0018] According to another aspect, the invention relates to an additive manufacturing process metallique by selective consolidation of a powder bed including a step of determining the oxygen concentration of said powder bed according to the invention, implemented at least once during said metal additive manufacturing process.

[0019] According to one embodiment of the metal additive manufacturing process according to the invention, the method for determining the oxygen concentration of the powder bed is applied at the beginning of the metal additive manufacturing process, during the injection of the inert gas into a manufacturing chamber, and / or at the end of the metal additive manufacturing process, during the cooling of a manufactured part.

[0020] According to one embodiment, the metal additive manufacturing process according to the invention includes a step of adapting process parameters according to the determined oxygen concentration value.

[0021] According to one embodiment, the metal additive manufacturing process according to the invention further includes a step of mixing the used powder with new powder when the oxygen concentration is above a predetermined threshold, the mixing step being carried out between two manufacturing of parts.

[0022] According to another aspect, the invention relates to a device for determining the oxygen concentration of a powder of a metallic material having a powder bed shape, comprising: - an image capture device configured to produce an image of at least a portion of said powder bed, said image comprising a set of pixels, each pixel having a color coded according to a colorimetric coding comprising three quantities, and - a first processing unit configured to determine the oxygen concentration of the powder from values ​​of said three quantities associated with pixels of the image, using a predefined calibration function, a function of said material, and relating said oxygen concentration and said three quantities.

[0023] According to another aspect, the invention relates to a computer program comprising instructions which lead the device according to the invention to execute the steps of the method for determining the oxygen concentration according to the invention.

[0024] According to yet another aspect, the invention relates to a metal additive manufacturing system by selective consolidation of a powder bed comprising: - a device for determining oxygen concentration according to the invention, - a reservoir designed to hold a substrate on which a part will be manufactured, - a device for spreading said powder on the surface of the powder bed, - a powder consolidation device, and - a second processing unit configured to control the implementation of metal additive manufacturing.

[0025] According to one embodiment, the image capture device comprises a scanner with flat plate connected to the spreading device.

[0026] According to one embodiment, the image capture device includes a high-resolution camera.

[0027] According to yet another aspect, the invention relates to a method for determining a calibration function for implementing the method for determining an oxygen concentration according to the invention, comprising the steps of: - to have a plurality of samples of said powder, each sample having a different and known oxygen concentration, - to produce an image of each sample, and determine an associated average color, a color being coded according to a colorimetric coding comprising three quantities, a color associated with an oxygen concentration value being called the calibration value, and - determine said calibration function by regression from said calibration data.

[0028] The following description presents several embodiments of the device of the invention: these examples are not limiting to the scope of the invention. These embodiments present both the essential features of the invention and additional features related to the embodiments considered.

[0029] The invention will be better understood and other features, objectives and advantages thereof will become apparent from the following detailed description and with reference to the accompanying drawings given by way of non-limiting examples and in which:

[0030] The [Fig.1] already cited illustrates a system implementing metallic additive manufacturing by selective consolidation of a powder bed.

[0031] Fig. 2 illustrates the method for determining the oxygen concentration of a powder having a powder bed shape according to the invention.

[0032] Figure 3 illustrates two applicable methods for implementing step B of the process according to the invention.

[0033] Figure 4 illustrates the first method for carrying out step B according to the invention.

[0034] Figure 5 illustrates the second method for carrying out step B according to the invention.

[0035] Figure 6 illustrates a first variant of the device according to the invention in which the image capture device comprises optics and a matrix detector.

[0036] Figure 7 illustrates a metal additive manufacturing system by selective consolidation of a powder bed according to the invention.

[0037] Figure 8 illustrates in more detail a metal additive manufacturing system by selective consolidation of a powder bed according to the invention.

[0038] Figure 9 illustrates the calibration data for the 22 samples produced and re presents the intensity of the three components R, G and B associated with the different concentrations of the samples.

[0039] Fig. 10 illustrates the color associated with 7 samples represented in the CIExy 1931 colorimetric space.

[0040] Fig. 11 illustrates the different oxygen concentration measurements carried out with method 1 and method 2 according to the invention, and with an ex situ method by fusion under inert gas.

[0041] Fig. 12 illustrates the oxygen concentration results obtained with methods 1 and 2 of the process according to the invention, the ex situ method by fusion under inert gas, and theoretical values ​​for the four samples. DETAILED DESCRIPTION OF THE INVENTION

[0042] The method 20 for determining the oxygen concentration Cox of a powder MP of a metallic material Mat having a powder bed shape PB according to the invention is illustrated [Fig. 2]. It is based on colorimetry and uses the color of the particles to determine the oxygen content of the powder, which, as described above, is a good indicator of powder quality degradation in metal additive manufacturing.

[0043] The mass concentration of oxygen Cox of a particle represents the amount of oxygen that makes up the composition of the particle, and is mainly present at its surface. It is typically measured in particles per million weight, wppm in Anglo-Saxon terminology.

[0044] During the high-temperature oxidation of metal powders, an oxide layer forms on the surface. Different layer thicknesses result from different heating conditions and lead to different oxygen contents. Colors are observable and obtained due to these different oxide film thicknesses and interference between light reflected by the film / metal interface and light reflected by the upper part of the oxide (film / air interface). It is therefore possible to correlate the color of a particle with its oxygen content.

[0045] The method according to the invention performs the analysis directly on the PB powder bed, that is to say on a layer of powder having a flat surface.

[0046] It comprises a first step A consisting of producing an image Im of at least a portion of the powder bed PB. The image comprises a set of pixels Pi i, the pixel index, and during image acquisition, a pixel of the image exhibits a color Coli coded according to a colorimetric coding comprising three quantities Gl, G2, G3. Due to visual trivariance, three numbers are sufficient for identifying a color.

[0047] According to one embodiment, the colorimetric coding is the RGB system (or RGB additive synthesis) and G1=R, G2=G and G3=B. This is done by adding light To reproduce different colors, most colors can be created / characterized by the additive synthesis of three light beams: red (R), green (G), and blue (B). To display / characterize a specific color, the relative importance of each of the three additive RGB primaries that contribute to its composition is determined. The absolute sum of these three colors produces white. This model is very common because it corresponds to the operation of color monitors. Each primary color oscillates, either as a percentage between 0% and 100% or as a value between 0 and 255 (8-bit encoding, 24 bits in total per color): a particular color is thus specified by indicating the contributions of each primary color. Having 256 shades of each primary color allows for the creation of 16.7 million colors (256 x 256 x 256). This encoding is also used when creating a color image of a scene with a detector equipped with color filters.With RGB encoding, a Pi pixel is associated with a triplet (Ri, Gi, Bi).

[0048] But any type of color coding is usable in the invention, for example (cyan, magenta, yellow) coding, or CIE XYZ coding which derives from RGB coding, or CIE U'V'W' coding, CIE L*a*b* coding.

[0049] In one embodiment, the method is implemented using the initial encoding of the image-capturing device, or in another embodiment, a transformation (typically a change of coordinate system) is performed to another encoding. The important point is to have information about the color of the pixels in the powder bed image PB, this information being encoded by a triplet of values.

[0050] In a second step B, the oxygen concentration of the powder is determined from values ​​of the three quantities associated with pixels of the image, using a predefined CFMat calibration function, a function of the Mat material, and relating the oxygen concentration Cox and the three quantities:

[0051] Cox = CFMat(Gl, G2, G3)

[0052] Let Pj be the pixels of the image that are used in the process according to the invention. The pixels Pj may correspond to all the pixels of the image or to a part of the pixels thereof.

[0053] The oxygen concentration Cox of the powder MP arranged in the form of a powder bed PB and imaged is thus determined from a set of triplets (G1j, G2j, G3j) associated with the pixels Pj.

[0054] The inventors have identified, after numerous experiments and research, CFMat calibration functions linking a given color to a unique oxygen concentration within the range of interest (bijective relationship between color and oxygen concentration). This range of interest is typically an oxygen mass concentration, measured in wppm, within the interval [200, 2500]. Furthermore, it has been demonstrated that color measurement performed by imaging devices of the This trade is compatible with a precise determination of oxygen concentration. Note that several calibration functions may correspond to a given material (Mat).

[0055] Very good results were obtained for measuring oxygen concentration. This is a remarkable and surprising result. The degradation mechanisms of powder in laser powder bed fusion are indeed very complex due to laser-matter interaction and the physical phenomena occurring in the melt pool, with convection currents and ejecta of liquid material, and the entrainment of particles near the laser by recoil pressures. In the recycled powder, colored particles are found, but also particles with surface oxide nodules, without continuous oxide films causing color (see, for example, the aforementioned publication by Delacroix et al.). It was therefore not obvious a priori that determining oxygen by colorimetry to control the quality of powders in additive manufacturing would be predictable and comparable to conventional measurements (see below).

[0056] The method 20 according to the invention takes advantage of the fact that it is possible to image the powder bed because it has a planar structure. It offers the advantage of performing in situ and non-contact characterization of the oxygen concentration by acquiring an image of the powder bed and determining the oxygen content of the powder layer using image colorimetry. It does not require handling the powder, is not applied to samples but directly to the powder bed, and can therefore be easily integrated into a process using a powder bed (see below).

[0057] Furthermore, its implementation requires few resources, as the calibration function is determined independently and stored in memory. The inventors have also demonstrated that the calibration function can be expressed as a first- or second-degree polynomial in three variables corresponding to the three quantities, and whose coefficients are a function of the material Mat (see below for an example of determining this calibration function).

[0058] According to a first variant, step A (creation of the image of the powder bed) is carried out in one go with a camera and according to a second variant detailed later step A is carried out by scanning with a flatbed scanner.

[0059] In the remainder of this exposition, the three RGB quantities will be used for color coding.

[0060] Step B of the process according to the invention can be carried out according to two methods illustrated [Fig.3].

[0061] According to a first method also described [Fig. 4], the calibration function is applied to obtain the oxygen concentration for each pixel, and then the average oxygen content of the bed is calculated by summing over all the individual pixels Pj. Thus, it is first determined in a step B1 and for a plurality of pixels Pj, the corresponding oxygen concentration Coxj, called pixel concentration, via the calibration function:

[0062] Coxj=FCMat(Rj, Gj, Bj)

[0063] Then, in a step B2, the oxygen concentration is determined from an average of the pixel Coxj concentrations.

[0064] Cox — Coxj

[0065] According to a second method also described [Fig.5], an average value (Rm, Gm, Bm) of each quantity is first determined in a step B' 1 from values ​​(Glj, G2j, G3j) of the quantities associated with a plurality of pixels of the image Pj.

[0066] Then, in a step B'2, the oxygen concentration is determined from the average values ​​of the three quantities via the calibration function:

[0067] Cox = FCMat(Rm, Gm, Bm).

[0068] The first method should theoretically be more robust because it applies the calibration function to each pixel, and the average of these Coxj concentrations can therefore be assumed to represent the physical value of the amount of oxygen present in the powder bed. The second method is faster because the image processing determining the average values ​​Rm, Gm, Bm over a set of pixels Pj is very simple and the calibration function is applied only once.

[0069] Given the diversity of colored particles present on the powder beds of degraded raw materials, it was not obvious a priori that the two methods would give similar results. Moreover, from a mathematical point of view, significant differences could have been expected when the correlation function used is not linear. However, the inventors compared the results obtained with the two methods, and the second method surprisingly gave results as good as the first method (see example below).

[0070] The fact that the process according to the invention operates in situ on a powder bed allows it to be easily integrated into a MAM-PBF process, an example of which is described in the prior art. In another aspect, the invention relates to a metal additive manufacturing process by selective consolidation of a powder bed, including a step for determining the oxygen concentration of the powder bed according to process 20 of the invention. Process 20 is implemented at least once during the MAM-PBF process.

[0071] According to one embodiment, the process 20 is applied at the beginning of the metal additive manufacturing process, during the injection of the inert gas into a manufacturing chamber, and / or at the end of the metal additive manufacturing process, during the cooling of a manufactured part.

[0072] According to one embodiment, step A of process 20 is carried out by scanning and simultaneously with a powder spreading step (see below).

[0073] Thanks to the process 20 according to the invention, there is a new opportunity to control the quality of powders directly online, in the machine, on samples directly used for manufacturing and totally representative.

[0074] Having in-situ oxygen concentration measurement available during the MAM-PBF process allows for adjustments that were previously impossible. In one embodiment, the MAM-PBF process includes a step for adapting the process parameters based on the determined oxygen concentration value. This involves, for example, modifying the laser power, the laser scanning speed, or the distance between two laser trajectories, all of which affect the local energy density applied to the material. More energy must be applied to a powder surrounded by an oxide layer to remove it by melting or vaporization.

[0075] Furthermore, it is now possible to avoid using poor-quality powders, i.e., those with an excessively high oxygen concentration. According to one embodiment, the metal additive manufacturing process includes a step of mixing the used powder with fresh powder when the oxygen concentration of the used powder exceeds a predetermined threshold C0. The mixing step is carried out between two part manufacturing runs.

[0076] According to another aspect, the invention relates to a device 10 for determining the oxygen concentration Cox of a powder MP of a metallic material Mat having a powder bed shape PB. The device 10 comprises an ICD image capture device configured to produce an image Im of at least a portion of the powder bed, the image Im comprising a set of pixels Pi, each pixel of the image having a color coded according to a colorimetric coding comprising three quantities (G1, G2, G3). The device 10 also comprises a first processing unit UT1 configured to determine the oxygen concentration of the powder from values ​​of the three quantities associated with pixels of the image, using a predefined calibration function CFMat, a function of the material Mat, and relating the oxygen concentration Cox and the three quantities.

[0077] A first variant of device 10 according to the invention is illustrated [Fig. 6] in which the ICD device comprises an Opt optic and an MD matrix detector. The image is captured in a single pass. Preferably, the image capture device comprises a high-resolution camera.

[0078] According to a second variant of device 10 according to the invention, the ICD device comprises a flatbed scanner operating an image capture of the powder bed by scanning.

[0079] According to another aspect, the invention relates to a 100 system for metal additive manufacturing by selective consolidation of a powder bed illustrated [Fig.7].

[0080] The system 100 includes a Tk reservoir intended to hold a Sub substrate on The system includes a powder spreading device (PSD) for manufacturing a part (Pa), a powder spreading device (CD), and a powder consolidation device (MD). Preferably, it also includes a vertical substrate displacement device (MD). The system also includes a device (10) according to the invention, illustrated in Fig. 7, in the second variant, the ICD device comprising a flatbed scanner (Scan). The system also includes a second processing unit (UT2) configured to control the implementation of metal additive manufacturing. Preferably, the first processing unit (UT1) is integrated into UT2.

[0081] According to one embodiment, the ICD image capture device of system 100 comprises a Scan flatbed scanner connected to the PSD spreading device. Image capture is then performed with a device integral to that used for powder spreading, which is advantageous in terms of machine integration. The addition of a scanner to an additive manufacturing machine has also been described, for example in US patent 10981225, for surface inspection.

[0082] A more detailed implementation of a system 100 with a Scan scanner is illustrated [Fig.8].

[0083] As a reminder, the principle of the scanner is as follows: a lamp, located on a moving block, scans the entire surface of the document / surface. The operation is performed in increments. It divides the document / surface into virtual lines, and it is the increment of the block's movement that determines the horizontal resolution of the scanner. A sensor receives the light reflected by the document / surface and defines the color of the dots that make up each line. Flatbed scanners are equipped with two technologies: CCD or CIS (Contact Image Sensor).

[0084] In a CCD scanner, the lamp emits white light, which is then reflected line by line via a series of mirrors. At the end of its path, the beam passes through a lens that concentrates the light rays and focuses them onto the CCD sensor, which is composed of arrays of photosensitive elements. To reproduce the color of the documents / surfaces, red, green, and blue filters alternately cover them. The sensor measures the amount of light received line by line and transforms it into an electrical charge, which is then converted into digital data.

[0085] In CIS technology, the light source consists of diodes (LEDs) emitting red, green, and blue light, and the sensors are arranged across the entire width of the scanner and move in tandem with the LEDs. A cylindrical lens focuses the light emitted by the LEDs onto the sensor in three colors. The LEDs, lens, and sensor are all part of the same device, one per pixel across the width.

[0086] Both technologies are compatible with the invention, and commercial scanners generally deliver colorimetric information per pixel coded at 1 at 255 on the three colors RG and B, but CIS technology is preferred as it offers better colorimetric performance.

[0087] Several resolutions have been tested and are functional. Preferably, a high resolution, typically 2400 dpi or even 4800 dpi, which allows for resolving the particle size, is preferred for a more accurate measurement of the powder bed color. Since the speed of the spreading device is relatively fast (e.g., 50 mm / s), in one embodiment, image capture is performed via the scanner attached to the spreading device as it returns to its initial position, at a speed compatible with the high resolution (e.g., 0.16 mm / s), and not during the actual spreading of the powder.

[0088] High resolution is preferred because it allows the particle size to be resolved, and the degradation of a powder bed is not homogeneous but consists of colored particles dispersed on the surface.

[0089] A method for determining the calibration function FCMat is now described, illustrated by an experimental example, for implementing the method 20 according to the invention.

[0090] First of all, it is necessary to have a plurality of samples of the powder, each sample having a different and known oxygen concentration.

[0091] Consider a powder of SS316L stainless steel. Various oxidations of this powder are carried out by subjecting the samples to different combinations of time and temperature in a furnace. The oxygen concentration is then measured ex situ using an inert gas fusion (IGF) method known to those skilled in the art. Visually, the color of the powder varies from gray to orange / brown at low concentrations, and as the concentration increases, the powder becomes pink and then blue.

[0092] Next, an image is taken of each sample, and the associated average color is determined according to a colorimetric coding comprising the three quantities (G1, G2, G3). A color associated with an oxygen concentration value is called the calibration data.

[0093] This measurement was performed on the samples using a Canon CIS scanner with a resolution of 4800 dpi, and the color was measured for all pixels of the recorded image. Then, an average value of R, G, and B was calculated from the values ​​measured for the different pixels.

[0094] Fig. 9 illustrates the calibration data for the 22 samples produced with oxygen concentrations within the range of interest [200, 2500] wppm, and represents the intensity of the three components R, G and B (each coded from 1 to 255) associated with the different concentrations of the samples.

[0095] By way of illustration, the color associated with 7 numbered samples from 1 to 7 chosen among the 22 (see Table I below) and was calculated and represented [Fig. 10] in the CIExy 1931 color space commonly used to represent colors.

[0096] Sample 1 2 3 4 5 6 7 Measured Cox (ppm mass) 475 1055 1347 1579 1707 1946 2458 Table I: Characterized samples

[0097] From the respective position of the 7 colors associated with the 7 samples of 316L stainless steel powder exhibiting increasing oxygen content, it is observed that on the range of interest of Cox the colors are very different, and vary from grey to orange then to blue passing through pink.

[0098] Finally, a calibration function is determined by regression from the calibration data. Preferably, a first- or second-degree polynomial regression is used, which have given good results.

[0099] The variables RG and B are respectively denoted x, y and z in the following equations.

[0100] According to a first example applied to the data in [Fig.9], the CFlMat calibration function was determined by regression considering a polynomial of the type (1, x, y, z, xy, xz, yz, x2, y2, z2) and we obtain:

[0101] CFlMat = 4036 - 49 x + 0.24 x2 + 45 y - 0.60 xy + 0.05 y2 - 18z + 0.3 xz + 0.36 yz - 0.40 z2

[0102] With this first regression, a coefficient of determination of cor is calculated. relation R2 = 0.99824

[0103] According to a second example applied to the data in [Fig.9], the CF2Mat calibration function was determined by regression considering a polynomial of the type (1, x, y, z, xy, xz, yz) and we obtain:

[0104] CF2Mat = 6117 - 95 X + 178 y - 0.77 xy - 127 z + 1.6 xz - 0.77 yz

[0105] With this second regression, we calculate a coefficient R2 = 0.99642

[0106] According to a third example applied to the data in [Fig.9], the CF3Mat calibration function was determined by regression considering a polynomial of the type (1, x, y, z, x2, y2, z2) and we obtain:

[0107] CF3M. = 2906 + 13 .x - 0.30 x2 - 2.5 y + 0.40 v2 - 6.22 z- 0.22 z2

[0108] With this third regression, we calculate a coefficient R2 = 0.99676

[0109] Given the R2 values ​​very close to 1, we observe that the three regressions give good results.

[0110] Using the CFlMat calibration function above, the process 20 according to the invention is tested. For this purpose, samples of 316L stainless steel powder are used. new (RO), and powders that have been used in a MAM-PBF system (by laser melting) and recycled several times: 1 time (RI), 5 times (R5), 10 times (RIO) and 15 times (R15). These recycling processes were studied in the aforementioned publication by Delacroix et al. and involve the laser melting of several parts on a platform, the recovery of all unconsolidated powder as a part, the sieving of this powder to remove the largest particles, and the reintroduction of this sieved powder into the machine for further manufacturing, without the addition of new powder.

[0111] During passage through the machine, the powder bed is scanned. Digital zooms on the scans highlight a heterogeneous structure of the particles with changes from particle to particle and show the multitude of oxidized particles of different colors in the scan of R15.

[0112] Preferably, the same image capture device should be used for determining the calibration function and for the subsequent characterization of the powder.

[0113] From the images obtained by scanning for each RO powder bed at R15, the oxygen concentration is determined using method 1 and method 2, and compared to measurements carried out ex situ by fusion under inert gas. Figure 11 illustrates the different Cox values ​​obtained.

[0114] It is observed that the Cox values ​​obtained with the process according to the invention successfully follow the trend towards increasing oxygen concentration as the powder is reused more and more.

[0115] A remarkable result is that the results obtained by the process according to the invention (methods 1 and 2) are perfectly in line with ex situ measurements. The results with methods 1 and 2 according to the invention are slightly higher than the ex situ results, up to 10-15 wppm, but remain practically always within the standard deviation of the chemical analysis values.

[0116] Another positive result is that the second method leads to results extremely close to those obtained with the first method (less than 5 wppm difference). Therefore, only the second method can be applied for the analysis of powder bed scans, as this approach provides almost instantaneous oxygen concentration results (code execution time of 70 ms for a 100 mm² area / 7 s for a 10 cm² area). The first method is nevertheless feasible and more rigorous.

[0117] As explained above, it was not obvious a priori that determining oxygen by colorimetry to control the quality of powders in additive manufacturing would be predictable and comparable to conventional measurements. Furthermore, the theoretical approach stating that color is solely determined by the thickness of the the film around the particle is certainly not totally accurate, and the fact that the process according to the invention gives an exact result of oxygen concentration with a highly recycled powder (RIO and R15) is in itself remarkable and surprising.

[0118] Furthermore, the oxidation during the process takes place in an atmosphere under an inert gas (generally argon, nitrogen, or helium), with very low partial pressures of oxygen and unknown and non-uniform temperatures experienced by the powders. It was therefore not obvious that the correlation between oxygen and color obtained via controlled oxidation in an air oven at fixed temperatures (for determining the calibration function) would allow for the retrieval of similar values ​​for colored particles due to oxidation during the laser powder bed fusion process.

[0119] To further test the robustness of the process according to the invention, samples of "artificially" degraded powder were also analyzed. Mixtures of fresh powder were prepared containing different fractions of oven-oxidized powders (5 and 10% by weight) with different oxygen levels, L1 at 1580 wppm and L2 at 2350 wppm. This yielded four samples: - Sample [L1-5%]: 95% new powder - 5% powder oxidized at the L1 level, - Sample [L1-10%]: 90% new powder - 10% powder oxidized at the L1 level, - Sample [L2-5%]: 95% new powder - 5% powder oxidized at the L2 level, and - Sample [L2-10%]: 90% new powder - 10% powder oxidized at the L2 level.

[0120] Figure 12 illustrates the Cox concentration results obtained with methods 1 and 2 of the process according to the invention, the ex situ method by melting under inert gas, and theoretical values ​​for the four samples. These theoretical values ​​represent the expected oxygen concentration of the powder mixtures based on the mass fractions and oxygen contents of the two constituents.

[0121] The inert gas fusion measurements of the four samples are in almost perfect agreement with the theoretically calculated values.

[0122] With regard to the results obtained with methods 1 and 2 according to the invention, both methods slightly overestimate the values ​​compared to those of the chemical analysis. The trends are still respected, and the results of the two methods are once again virtually identical. For a given fraction of oxidized particles in the coatings, the overestimation of Cox is more pronounced for the LL samples. These are composed of light orange particles, while the L2 samples contain blue particles. It is observed that the edges of the Colored particles appear darker in the acquired images, potentially leading to overestimation. Since the L2 oxidized particles are already quite dark, the edge effect is less pronounced, with fewer color variations, which could explain the smaller discrepancies measured for the L2 mixtures. Furthermore, for a given level, the difference is greater for samples with 10% colored particles. The increased number of colored particles induces a more pronounced edge effect, resulting in larger discrepancies. It is nonetheless remarkable, and also quite surprising, that the method according to the invention allows for a Cox estimation so close to reality for a powder comprising a mixture of different oxidized particles.

Claims

Demands

1. A method (20) for determining the oxygen concentration (Cox) of a powder (MP) of a metallic material (Mat) having a powder bed shape (PB), said method comprising the steps of: - A) Making an image (Im) of at least a part of said powder bed, said image comprising a set of pixels (P), a pixel having a color coded according to a colorimetric coding comprising three quantities (G1, G2, G3), - B) Determining the oxygen concentration of the powder from values ​​of said three quantities associated with pixels of the image, using a predefined calibration function (CFMat), a function of said material, and relating said oxygen concentration and said three quantities.

2. A method according to the preceding claim in which the colorimetric coding is the RGB system, the three quantities being designated R, G and B.

3. A method according to any one of the preceding claims wherein the calibration function is a first or second degree polynomial in three variables corresponding to the three quantities, and whose coefficients are a function of said material.

4. A method according to any one of the preceding claims wherein step A is carried out in one step with a camera.

5. A method according to any one of the preceding claims wherein step A is carried out by scanning with a flatbed scanner.

6. A method according to any one of the preceding claims wherein step B comprises the substeps of: - B1) Determining for a plurality of pixels of the image an oxygen concentration (Coxj), called pixel concentration, via the calibration function, and - B2) Determining the oxygen concentration from an average of said pixel concentrations.

7. A method according to any one of claims 1 to 5, wherein step B comprises the substeps of: - B' 1) Determine an average value (Glm, G2m, G3m) of each quantity from values ​​(Glj, G2j, G3j) of said quantities associated with a plurality of pixels of the image, and - B'2) Determine the oxygen concentration from said average values ​​of the three quantities via the calibration function.

8. A metal additive manufacturing process by selective consolidation of a powder bed including a step of determining the oxygen concentration of said powder bed according to any one of claims 1 to 7 implemented at least once during said metal additive manufacturing process.

9. A metal additive manufacturing process according to the preceding claim, wherein the method for determining the oxygen concentration of the powder bed according to any one of claims 1 to 7 is applied at the beginning of the metal additive manufacturing process, during the injection of the inert gas into a manufacturing chamber, and / or at the end of the metal additive manufacturing process, during the cooling of a manufactured part.

10. A metal additive manufacturing process according to any one of claims 8 or 9 comprising a step of adapting process parameters according to the determined oxygen concentration value.

11. A metal additive manufacturing process according to any one of claims 8 to 10 further comprising a step of mixing the used powder with new powder when the oxygen concentration is above a predetermined threshold, said mixing step being carried out between two part manufacturing runs.

12. Device (10) for determining the oxygen concentration (Cox) of a powder (MP) of a metallic material (Mat) having a powder bed shape (PB), comprising: - an image capture device (ICD) configured to produce an image (Im) of at least a portion of said powder bed, said image comprising a set of pixels (P), a pixel having a color coded according to a colorimetric coding comprising three quantities (G1, G2, G3), and - a first processing unit (UT1) configured to determine the oxygen concentration of the powder from values ​​of said three quantities associated with pixels of the image, using a predefined calibration function (CFMat), function of said material, and relating said oxygen concentration and said three quantities.

13. A computer program comprising instructions that cause the device of claim 12 to perform the steps of the process according to any one of claims 1 to 11.

14. System (100) for metal additive manufacturing by selective consolidation of a powder bed comprising: - a device (10) for determining the oxygen concentration according to claim 12, - a reservoir (Tk) for receiving a substrate on which a part will be manufactured, - a device (PSD) for spreading said powder on the surface of the powder bed, - a device (CD) for consolidating the powder, and - a second processing unit (UT2) configured to control the implementation of metal additive manufacturing.

15. Metal additive manufacturing system according to the preceding claim wherein the image capture device comprises a flatbed scanner (Scan) connected to the spreading device (PSD).

16. Metal additive manufacturing system according to claim 14 wherein the image capture device comprises a high-resolution camera.

17. Method of determining a calibration function for implementing the process according to any one of claims 1 to 7 comprising the steps of: - having a plurality of samples of said powder, each sample having a different and known oxygen concentration, - taking an image of each sample, and determining an associated average color, a color being coded according to a colorimetric coding comprising three quantities (G1, G2, G3), a color associated with an oxygen concentration value being called calibration data, and - determining said calibration function by regression from said calibration data. 20

Images