METHOD FOR DETERMINING BEAM ASSIGNMENT DATA FOR REGIONS IN POWDER METAL FILLING EQUIPMENT USING BEAM METHODS

Abstract

The invention relates specifically to a method for determining beam-assigned data for a region in a beam-based powder fusion smelting apparatus, which includes: - s2) for a set of gradually solidified dashed lines by beam irradiation, the distribution of the set of dashed lines into distinct subsets, each subset comprising a number of dashed lines and forming a region (R1 – R8) processed by one or more beams of the apparatus, such distribution includes, for at least some regions, the grouping of a number of dashed lines, each untruncated, the dashed lines placed side by side and distributed successively along a certain direction perpendicular to the dashed lines, - s3) the determination of beam-assigned data for the region.

Description

Method for determining beam-to-region assignment data for a powder-bed beam fusion manufacturing device

[0001] The technical field is that of additive manufacturing, more specifically powder-bed beam fusion additive manufacturing.

[0002] Additive manufacturing techniques are undergoing strong growth, due to the considerable manufacturing freedom they allow. They enable to manufacture parts having complex shapes, for instance hollow parts, all in one piece. And an additive manufacturing device can build many different parts with no modification of the device itself and no long prototyping phase.

[0003] Powder-bed beam fusion is an additive manufacturing technique in which a raw material powder is applied in layers to a support stage which is vertically movable by means of a moving device. The material powder is then solidified, layer by layer, to gradually build the part. For each layer, different zones of the layer are progressively solidified by melting or sintering the powder locally using a beam, such as a laser beam or an electron beam.

[0004] Though highly versatile, such a manufacturing technique is usually slow, with a rather low throughput compared to other manufacturing techniques like stamping or moulding.

[0005] It is thus desirable to increase the throughput of such techniques.

[0006] A solution for this is to use a powder-bed beam fusion device that emits multiple beams, so that different zones of a layer can be solidified simultaneously. Document W02020160103 describes such a powder-bed beam fusion device, in which multiple laser beams are employed. In the method described in this document, a section to be solidified is divided in many different small regions. These regions are then each assigned to one of the lasers of the device, and a laser-to-region sequence is determined for the manufacturing. This assignment and scheduling are determined so as minimize a total process time required for building the section of the part considered (in other words, for solidifying a given “slice" of the part).

[0007] Besides, different computer programs, enabling to determine a beam-to-region sequence for piloting the lasers of a powder-bed laser fusion device, are also commercially available (usually provided with the device in question).

[0008] One goal of the instant technology is to improve further these methods and programs, in particular in terms of manufacturing throughput.Summary

[0009] In this context, a method according to claim 1 , for determining beam-to-region assignment data for a powder-bed beam fusion device, is provided.

[0010] In terms of geometric features, different levels of detail can be considered when planning manufacturing a part using powder-bed beam fusion. At a high-level point of view, it is the geometry (in 3 dimensions) of the surface that delineates the part (geometry of its outer surface) that is considered. At a medium-level point of view, it is the geometry (in 2 dimensions) of the edge(s) of a section of the part (section corresponding to one layer of power, to be fused) that is considered (in other words, it is the geometry of the boundaries that delineate this section of the part, laterally). And at the most basic level (somehow a microscopic level), it is the detailed structure of the vectors or hatch-lines to be scanned by the beam(s) which is considered.

[0011] Remarkably, in this method, the section of the part is divided into different regions, to be beam-scanned, based directly on the structure of the ensemble of hatch-lines to be scanned, that is, based on the thinner level of details, in the hierarchy above presented. This is very different from dividing the section of the part based directly on the medium-level of details (that is based on an outer geometry of this section), as it starts from the smaller subdivisions (the hatches) and then groups them, instead of dividing the section, at a macroscopic level. Partitioning the section into regions with the ensemble of hatch-lines as a basis lead to more precise and more efficient partitioning (as the allocation of a region to the beam or beams is achieved based on what the beam is expected to scan - namely based on the hatch-lines themselves). Though leading to higher performances, this portioning is harder to achieve as the details taken into account are very thin and numerous (a typical scanning data may comprise thousands of hatch-lines). To overcome the complexity arising from starting from the very detailed structure of the scans to be achieved, representing the ensemble of hatch-lines by the graph defined in claim 1 is very efficient; indeed, this graph summarizes the relative positioning of the hatch-lines that matters for linking scans, and allows, by dividing the graph using reasonable fast procedures, to form appropriate regions. Besides, using the graph that is defined in claim 1 allows for identifying efficiently hierarchical structures in the ensemble of hatch-lines (identifying hatches that belong to a same class, from a mathematical point of view), which facilitates the transition from a microscopic description to a macroscopic description of the section of the part.

[0012] In this method, at least some of the regions, possibly most of them (or even all of them) may be defined, in step s2 (step of partitioning, or in other words of segmenting the ensemble of hatch-lines), so that the hatch-lines they gather are untruncated, and are located, in each of these regions, one aside the other, laterally adjacent to each other, one after the other successively. By untruncated, it is meant that the hatch-line, as specified in the acquired data (specified for instance by a start point and an end point), is allocated to a single region, not to two or more regions (with different parts of the initial hatch-line allocated to different regions to be scanned, the hatch-line being thus cut into parts, and, in each region, truncated).

[0013] So, in this method, the regions are defined based on the hatch-lines themselves, instead of corresponding to arbitrary geometrical areas dividing the section to be built, regardless of the hatch-lines organisation (like in document W02020160103, fig. 4, for instance). In particular, this method avoids fractioning (somehow cutting) a continuous hatchline into different regions (which would be detrimental in terms of production efficiency). Here, when a hatch-line belongs to the region, the beam assigned to this region can follow the entire hatch-line, from end to end.

[0014] Besides, as the regions in question each gather a set of latterly adjacent hatch-lines, all these hatch-lines can be scanned successively by a same beam, with no dead-time (no idle time; that is no time during which the beam is off), when these hatch-lines are scanned in the so-called “Zig-Zag scan” (for which one hatch-line is scanned in one direction, then the subsequent adjacent hatch-line is scanned in the opposite direction and so on, the beam moving back and forth). In a way, it is like if the ensemble of hatch-lines in such a region can be drawn with a pencil without lifting the pencil from the leaf of paper.

[0015] Forming at least some of the regions, so that they each gather a set of latterly adjacent, untruncated hatch-lines is favourable in terms of building quality (strength of the part, for instance) and productivity (reduces idle times, increases the build rate). In step s3, said characteristic of the process of manufacturing (manufacturing according to the beam-to-region data), which is optimized by an adequate beam-to-region assignment, may be a total idle time of the manufacturing process, or the total process time, or a homogeneity of the temperature within the layer of material, or a mix thereof. More generally, said characteristic of the process of manufacturing is a performance indicator for the process, related to quality or productivity performances. In this regard, the inventors have observed that the commercially available softwares mentioned above often lead to substantial total idle times, typically of the order of one third (or even more) of the total process time. It is thus highly desirable to reduce the beam or beams idle time, as achieved thanks to the instant method.

[0016] The hatch-lines may be distributed within each region, one after the other successively, from a first hatch-line of the region to one or more last hatch-lines of the region.

[0017] The first hatch-line of the region can be:- a section-edge hatch-line (like hatch-lines L1 , L3, L5 or L7 of figure 7), that is a hatchline which, on one side, has no laterally adjacent hatch-line (in other words, a hatchline that is located on the edge of the section to built, that is completely superficial); or a frontier hatch-line (like the hatch-lines Lf, in figure 7 or 17), that is a hatch-line having, on one side at least, more than one hatch-line (for instance two hatch-lines, like hatchlines La and Lb, in figure 7) that are laterally adjacent to the frontier hatch-line and that are on a same side of the frontier hatch-line,

[0018] The last hatch-line, or one of the last hatch-lines of the region considered can also be such a section-edge hatch-line or frontier hatch-line.

[0019] When the section of the part divides into two or more branches (see figure 7 for instance), these two or more branches cannot be scanned continuously (i.e.: with no switch off), by a single beam. In other words, one of the branches of the section would have to be chosen, to keep on the scan. It is thus beneficial to gather into different regions the hatch-lines that are located, respectively, on the different branches of the section. In practice, the hatchline which is at the frontier between such branches is a frontier hatch-line (see for instance the two hatch-lines Lf, in figure 17).

[0020] Delineating a region by such a frontier hatch-line thus enables to delineate the region at a point beyond which a continuous scanning of the whole section is not possible. It thus contributes to reducing the total idle time. Just as delineating a region by a section-edge hatchline.

[0021] Besides, the inventors have observed that, on the edges of the section, there may be series of short or even very short hatch-lines, located in the extension of each other, one after the other, just as if they belonged to a same virtually continued hatch-line (like hatch-lines L3, L5 and L7 of figure 7). Such hatch-lines can thus be conveniently scanned one after the other, in the continuity of one another using a same beam (the time during with this beam being off, between two hatch-lines, being small). Indeed, they are somehow located at the same level, among the different levels of hatch-lines scanned successively by the beam. And it is thus beneficial to gather such a series of “fringe” hatch-lines in a same region (rather than in distinct regions).

[0022] To this end, step s2 (step of partitioning) may comprise grouping together, in one of the regions, two (or possibly more) of the hatch-lines (for instance hatch-lines L3, L5 and L7 of figure 7), located in the extension of each other, one after the other in a direction parallel to the hatch-lines, when at least one of the two hatch-lines has, on one side, no laterally adjacent hatch-line (i.e.: is a section-edge hatch-line).

[0023] Moreover, grouping such a series of “fringe” hatch-lines in a same region reduces the total number of regions obtained at the end of step s2, which usually makes step s3 (step of beam-to-region assignment optimization) faster.

[0024] The method according to the instant technology may comprise one or more additional features, defined in claims 2 to 13, considered alone or in combination.

[0025] The instant technology also concerns a manufacturing method according to claim 14, a calculator according to claim 15, a manufacturing device according to claim 16 and a computer program according to claim 17.

[0026] The instant technology will now be described in more detail and illustrated by examples without introducing limitations, with reference to the appended figures.

[0027] Figure 1 schematically represents a powder-bed beam fusion device for additive manufacturing, viewed from the side.

[0028] Figure 2 schematically represents part of the device of figure 1 , viewed from above.

[0029] Figure 3 schematically represents, as a block diagram, steps of a method for determining beam-to-region assignment data for the device of figure 1 .

[0030] Figure 4 schematically represents a step of partitioning of the method of figure 3, in more detail.

[0031] Figure 5 is a top view of an ensemble of hatch-lines, for the beams of the device of figure 1 to follow to gradually solidify zones of a powder layer, in order to build a section of a part, in this case a complex part with many hollow zones and thin membranes intersecting one another.

[0032] Figure 6 is a detail view, from above, of a zone of the ensemble of hatch-lines of figure 5 (said zone being delineated by a dashed rectangle, in figure 5).

[0033] Figure 7 is similar to figure 6 but with a higher magnification.

[0034] Figure 8 shows partially a graph representation of the ensemble of hatch-lines of figure 5.

[0035] Figure 9 shows partially a condensed-graph obtained by condensing together some of the nodes of the graph of figure 8.

[0036] Figures 10 to 14 show partially the graph of figure 8, at different stages of a condensation operation.

[0037] Figures 1 1 B to 14B show further steps of the condensation operation, carried on the dual graph associated to the graph of figure 8.

[0038] Figure 15 partially represents the condensed-graph and highlight specific nodes of the condensed-graph, namely leaf and fork nodes.

[0039] Figure 16 partially represents the condensed-graph after its partitioning.

[0040] Figure 17 is similar to figure 7 but the regions, obtained after partitioning said ensemble of hatch-lines, are visible in figure 17 (they are represented respectively in different shades of grey).

[0041] Figure 18 partially represents a regions-graph wherein each of said regions is represented by a graph node.

[0042] Figure 19 is a perspective view of stamping inserts built by powder-bed beam fusion manufacturing, according to beam-to-region assignment data determined using the method of figure 3.

[0043] Figure 20 is a top view of a section of the stamping inserts of figure 19, showing a partitioning of the section into regions, this partitioning being obtained according to the method of figure 3.

[0044] Figure 21 shows the same section as in figure 20, but with a partitioning of the section into regions obtained using a prior-art software.

[0045] Figure 22 is a perspective view of nozzles, to be built by powder-bed beam fusion manufacturing.

[0046] Figure 23 is a top view of a section of one of nozzles of figure 22, showing a partitioning of the section into regions obtained according to the method of figure 3.

[0047] Figure 24 shows the same section as in figure 23, but with a partitioning of the section into regions obtained using said prior-art software.

[0048] In the following, the beam or beams are laser beam(s) (also designated as lasers’). Still, the instant technology could also be implemented using one or more particulate beams such as electron beam(s), instead of laser beam(s), for melting or sintering the material powder. Regarding the material powder, it could be a metal powder, a ceramic powder, a plastic powder, or another kind of powder.

[0049] In the following description, a powder-bed beam fusion device 1.1 is described first, with reference to figures 1 and 2. An exemplary embodiment of a method for determining beam-to-region assignment data for such a device is presented then, with reference to figures 3 to 18. Some test results are then presented with reference to figures 19 to 24, for two examples of parts produced by powder-bed beam fusion manufacturing.

[0050] Figure 1 schematically represents the powder-bed beam fusion device 1.1 , which comprises:- a support 1 .2 such as a horizontal tray, on which layers of raw material powder 1 .9, 1 .8, 1 .7 are successively deposited to produce a three-dimensional part P; the support 1 .2 is vertically movable by means of a moving device;- a dispensing device 1 .3 for depositing the powder layers 1 .9, 1 .8, 1 .7 onto the support 1.2; the dispensing device 1.3 comprises for example a wiper and / or a roller for distributing the powder and for leveling each new powder layer;- a beam source 1 .4 suitable to emit a number N of beam(s), here laser beam(s), N being equal to or above 1 , for totally or partially melting the last powder layer to have been deposited (i.e.: the top, surface layer 1.7); the respective direction of each beam emitted by the beam source 1.4 is electrically controllable (for instance by a galvanometric mirror), so it can scan the top powder layer 1 .7 to solidify selectively such or such zones of that layer;- a controller 1 .5, operatively connected to actuators of the powder-bed beam fusion device 1.1 , in particular to the beam source 1.4, to the movable device and to the dispensing device 1.3.

[0051] In the example presented in more detail below, the beam source 1 .4 emits four beams, B1 , B2, B3, B4. Still, the instant method can be applied to a powder-bed beam fusion device emitting an arbitrary number of beam(s) (including the case of a single beam powder-bed beam fusion device).

[0052] The manufacturing of the part P is carried-on layer by layer. One powder layer is deposited and some zones of this layer are solidified by the beams. A subsequent powder layer is then deposited on top of the previous powder layer, and some zones of this subsequent powder layer are then solidified by the beams (according to the shape of the part to be built) and so on. Here, before depositing each new powder layer, the support 1.2 is moved incrementally downwards. For each powder layer, the ensemble of zones that have been solidified form a section of the part P.

[0053] Figure 2 represents schematically the top powder layer 1.7 from above. The direction of a gas flow, flowing over the powder layer to remove smoke and particulates caused by powder melting, is represented by an arrow F.

[0054] The section S of the part is solidified progressively, line by line, by scanning the top powder layer 1.7 with the beams B1 - B4. To this end, each beam follows a path, on the surface of the top powder layer 1.7. In practice, an ensemble E of hatch-lines L, distributed over the section S and covering entirely or almost entirely this section, is defined (see figures 5 to 7). These hatch-lines L are each followed (in other words, each scanned) by one of the beams, to solidify the area extending along said hatch-line. In practice, each hatch line may be represented by a collection of segments each defined by two points Mi (being the start) and Mi+i(being the end), each segment being rectilinear. Each hatch line can be rectilinear, or wavy, or present a triangular or other pattern periodically repeated depending on the aggrupation of the segments composing the hatch line. A rectilinear hatch-line, for instance, may be composed of just one segment, while a sinusoidal hatch-line may be composed of many successive segments grouped together to form the hatch-line.

[0055] In the example represented in the figures, the hatch-lines are sinusoidal, as it helps presenting easily the influence of “fringe” hatch lines. Still, the instant method can be applied to any type of scanning (with either rectilinear or way hatch-lines, and using a Zig-Zag, Zig-Zig or other scanning strategy).

[0056] Anyhow, each hatch-line extends in a given direction (which corresponds to an average, straight line on which the hatch-line is centered, should the hatch-line not be rectilinear), this direction being parallel to fixed hatching direction Dh (see figure 7 for instance). So, the hatch-lines extend parallel to each other (as they each extend parallel to the hatchingdirection Dh). Each hatch-line extends from a first end to a last end (see the ends E1 and E2 of the hatch-line L in figure 7, for instance). The hatch-lines are separated one from each other, laterally, by a given pitch, for instance a constant pitch (whose value may be from 0.01 to 0.5 mm, for instance).

[0057] In practice, the hatching direction Dh is different from one layer to the other. There is an angular shift of the hatching direction from one layer to the other (as it is favorable for the strength of the part). And so, the ensemble of hatch-lines to be scanned, to solidify one section of the part, is different from one layer to the other (from one section to the other), here (even if the part has a constant section).

[0058] Each section to be solidified is subdivided, or in other words segmented into regions, each assigned to one the beams B1 - B4 (see for instance the regions R1 - R8 in figure 17, or the regions represented in figures 21 and 24). Each region is processed by the beam assigned (i.e. allocated) to it: within this region, the powder layer is (totally or partially) melted by the beam allocated to this region. Each region gathers a number of the hatch-lines of the ensemble of hatch-lines E (see figure 17).

[0059] So, each beam is assigned a set of regions, to be processed by that beam. More specifically, each beam is assigned a sequence of regions, to be processed successively, in a given order, by the beam considered.

[0060] Here, the beam-to-region assignment data specifies:- for each of said regions, which hatch-lines belong to the region considered, and- for each of the beams, which regions are to be processed by the beam considered, and in which order.

[0061] The above-mentioned controller 1.5 is configured for controlling the actuators of the powder-bed beam fusion device 1.1 , in particular the beam source 1.4, according to manufacturing instructions, in order to build the three-dimensional part P. These manufacturing instructions comprise, for each layer 1 .7 - 1 .9 to be processed, the beam-to-region assignment data associated to the layer considered. The manufacturing instructions may also comprise information specifying the positioning of the different hatch-lines, that is the positioning and extent of each hatch-line of the ensemble E. This positioning may be specified for instance by providing 2-dimensional coordinates (x-y coordinates), both for the first end and for the last end of each hatch-line.

[0062] The controller 1.5 may be configured to receive the manufacturing instructions from a calculator 1.6. The calculator 1.6 comprises at least a processor and a memory. It may take the form of a stand-alone computer, electronic unit or server. But it could also be implemented in a distributed manner (somehow “virtually”), using so-called “cloud” resources (computing and storing resources distributed among distinct physical systems in a network, possibly located at different places). The calculator 1 .6 and the controller 1 .5 may be distinct from eachother, or may be implemented as a single electronic device configured for planning and controlling the manufacturing process.

[0063] The calculator 1.6 is configured, more precisely programmed to execute the following steps (figure 3): s1 ) Acquiring data specifying the ensemble E of hatch-lines L for the beams B1 - B4 to follow to gradually solidify zones of the top powder layer 1 .7, to build the section S of the part,- s2) partitioning the ensemble E of hatch-lines into a plurality of distinct subsets, each gathering some of the hatch-lines and forming one of the above-mentioned regions, to be processed by one of the beams (B1 -B4),- s3) Determining the beam-to-region assignment data, which specify, for each of the beams B1 -B4, the sequence of regions to be processed by the beam considered, among said regions.

[0064] The method may also comprise outputting the beam-to-region assignment data, for instance to transmit it to the controller 1 .5, or to a manufacturing database. The calculator 1 .6 may be configured for controlling the powder-bed beam fusion device 1.1 so as to manufacture said section of the part, according to the beam-to-region assignment data.

[0065] As represented in figure 3, the beam-to-region assignment data may be transmitted to the controller 1 .5, which then controls the actuators of the powder-bed beam fusion device 1.1 , in step s4, to build the section S according to the beam-to-region assignment data. As above mentioned, the controller 1 .5 may also receive other manufacturing instructions for building the section S in question, in addition to the beam-to-region assignment data. It may receive beam- to-region assignment data for multiple sections of the part, to be built successively, for instance for all the successive sections of the part P (in which case the whole part is built, in step s4).

[0066] In step s1 , the data specifying the ensemble E of hatch-lines L, acquired by the calculator 1 .6 (for instance by loading a corresponding file), may be the information specifying the positioning of each hatch-line of this ensemble. Alternatively, these data may specify the start point and end point of each of the above-mentioned segments.

[0067] Alternatively, the data specifying the ensemble E of hatch-lines L may specify the start and end point of each of the elementary segments forming the hatch-lines. These data may also take the form of a specification of the contour of the section S (or even a specification of the geometry of the overall part P), supplemented by the direction Dh of the hatch-lines and their lateral spacing (that is the hatching pitch). In this last case, the calculator may then compute the positioning and extent of each hatch-line, based on this information.

[0068] Remarkably, step s2 comprises, for at least some of the regions (possibly for most of them, or even all of them):s22) grouping some of the hatch-lines L of the ensemble E to form the region considered, said hatch lines being each untruncated, being located one aside the other, and being distributed one after the other successively along a given direction Ds (figure 7) perpendicular to the hatch-lines.

[0069] As explained in the section entitled “summary”, defining the regions in this way, based on the hatch-lines organisation, reduces the total process time required to build the section S, as it reduces the beams idle times. This hatch-lines based segmentation of the section S is clearly visible in figure 17, for instance.

[0070] In step s3, each beam is assigned some of the regions, and is assigned an order in which processing the regions in question, using an optimization procedure configured: to minimize a total idle time, representative of the sum of the jump times for the one or more beams, each jump time being a time required, for one of the beams, to go from one of the regions of the sequence of regions to be processed by that beam, to the next region in said sequence;- or, directly, to minimize a total process time for building said section (this total process time taking into account the jump times, as well as elemental process times, required to processing, respectively, such or such region).

[0071] The total idle time may be equal to the sum of the different jump times, for the beam- to-region assignment considered. Still, total idle time may also take into account switch-on and / or switch off delays. Indeed, when a beam goes from one region to another distant region, it is switched off during this movement. Switching off the beam, and then on again requires a given extra time to be added to the corresponding, geometric-based jump time. The total idle may thus be calculated as a sum of adjusted jump times, each adjusted jump time being the sum of a (geometric-based) jump time and the switch-on and / or switch off times.

[0072] It is noted that reducing the total idle time enables in fact to reduce the total process time. Indeed, minimizing the time during which the beams are off means maximizing the total time during which the beams are on, and thus maximizing the building rate (the productivity) with which the section S is built, thus minimizing the total time required to build that section.

[0073] Steps s2 and s3 are now described in more detail.Step s2

[0074] Below, some general aspects of the segmentation of the section S in different regions are presented first, and a way to implement this segmentation using a graph representation (figure 8) is presented then in more detail.

[0075] The segmentation of the section S in different regions may be based, inter alia, on the two following criteria:- a) each region is a compact group of laterally adjacent hatch-lines, each untruncated (group that, in particular, may be scanned by a single beam with no interruption oralmost no interruption of the beam, if scanned in the Zig-Zag mode, except possibly for a few fringe hatch-lines),- b) hatch-lines that are on the edge of the section of the part (like hatch-lines L3, L5, L7 in figure 7), or close to this edge (like hatch-lines L2, L4, L6 in figure 7), when they are located in the extension of each other (one after the other in the hatching direction Dh, just as if they belonged to a same, virtually continued hatch-line), are grouped in the same region.

[0076] As explained in the “summary” section, a segmentation carried on according to criteria a) and b) enables to reduce the total idle time of the beams. It is noted that, in the instant description, laterally adjacent means immediately laterally adjacent (i.e.: with no intermediary hatch-line between).

[0077] Regarding criteria a), such compact groups the hatch-lines can each be scanned by a single beam with no interruption or almost no interruption of the beam, if scanned in the Zig- Zag mode (except possibly for a few fringe hatch-lines). And even if the hatch-lines in a same region are scanned in the so-called “Zig-Zig scan” (which corresponds to a raster-scan of the ensemble of hatch-lines, all scanned in the same direction, the beam being turned off when moving back to scan again a new hatch-line), instead of a “Zig-Zag scan”, the beam dead-time is minimized, as the displacements with the beam off are limited to what is just necessary to move the beam back and scan again a hatch-line

[0078] Regarding criteria b), it is noted that some hatch-lines, like hatch-line L and L8 in figure 7, may be located in the extension of each other (just as if they belonged to a same, virtually continued hatch-line), while it is not desirable to group them in the same region as the two hatch-lines in question belong to two different compact groups of hatch-lines. This is taken into account in the instant method (indeed, regarding L and L8, for instance, they are allocated to two different regions, when using the instant method). To this end, criteria b), of condensing together and grouping in a same region hatch-lines that are in the extension of each other, is applied on the condition that at least one of the hatch-lines in question (for instance all, or all but one of the hatch-lines in question) are fringe hatch-lines, that is hatch lines that are on the edge of the section of the part, or close to this edge.

[0079] By fringe hatch-line, it is meant more specifically:- a section-edge hatch-line (that is, a hatch-line with, on one side, no laterally adjacent hatch-line), or a hatch-line such that the number of hatch-lines, between the hatch-line considered and a section-edge hatch-line, is from 0 to a given maximum number. This maximum number is for instance equal to n0-2, n0being a threshold defined further below.

[0080] Before grouping the different hatch-lines into different regions, the step of partitioning, s2, may comprise an initial sub-step of grouping together the elementary segments above mentioned, so as to form the hatch-lines themselves.

[0081] Remarkably, in the embodiment presented here with reference to the figures, step s2 comprises a step s20 of building a graph G representing the ensemble E of hatch-lines (see figure 8), in which:- each hatch-line of the ensemble E is represented by a node of the graph, and- for any couple of two hatch-lines that are laterally adjacent to each other, the two associated nodes of the graph are linked by an edge of the graph.

[0082] In practice, graph G, and any other of the graphs below mentioned (Gc, Gr, G*) may take the form of (or, in other words, be represented by) an adjacency matrix.

[0083] The partition of the ensemble E of hatch-lines into different regions is then carried on by dividing this graph G, possibly preliminary condensed in the form of a condensed-graph Gc (see figure 9). Using this graph representation enables to partition the ensemble of hatch-lines according (inter alia) to the above-mentioned criteria, in an efficient manner from a computational point of view. Indeed, the possibility to scan continuously one hatch-line after another, during a beam scan, corresponds, on the graph G, to the possibility to go directly from one node to a next one, on the graph, following the graph edge. Besides, on this graph, hatchlines corresponding to limits of continuously scannable group of hatch-lines correspond to fork nodes (e.g.: bifurcation nodes) that are easy to identify.

[0084] In the embodiment corresponding to figures 9 to 16, step s2 comprises, after step s20:- a step of graph condensation s21 (see figure 4), during which the condensed-graph Gc is determined from graph G by merging some nodes of the graph, associated to fringe hatch-lines, onto other nodes of the graph (when the corresponding hatch-lines are located in the extension of each other), and then- a step s22 of partitioning the condensed-graph Gc.

[0085] Steps s21 and s22 are now described in more detail.

[0086] First, it is noted that a section-edge hatch-line, like L1 , L3, L5 or L8 in figure 7, is represented in the graph G by a leaf node, that is a node located at an end of a branch of the graph G, like nodes 1 , 3, 5, 8 of figure 8. In figure 8, the nodes 1 - 9 and 12 represent respectively the hatch-lines L1 - L9 and L12 of figure 7.

[0087] Regarding frontier hatch-lines, they are represented in the graph G by fork nodes, that is nodes with three or more edges connected to that node, like nodes 9, 12 or 13 of figure 8, and like the nodes labelled as FN in figure 15. step of graph condensation s21

[0088] During the step of graph-condensation s21 , when a node belongs to a low-length end branch (like end branches Br1 - Br4 of figure 8), it is merged onto another node of graph ifthese two nodes represent two hatch-lines that are in the extension of each other. This merging may be done, like here, on the further condition that said another node has a “deepness” smaller than or equal to a given limit, which is n0, here.

[0089] The deepness in question is the number of edges between a given node and the closest fork node. For instance, in figure 8, the node 1 has a deepness of 1 , the node 7 has de deepness of 2 and the node n (which represents the hatch-line L of figure 8) has a deepness of 3.

[0090] An end branch of the graph is a part of the graph extending from a leaf node to a common node associated to that leaf node, said common node being the first fork node encountered on the graph when starting from the leaf node considered and following the graph. The common node is not included in the end branch in question. A low-length end branch is an end branch whose total number of nodes is below or equal to a given, maximum number of nodes, here n0. n0is for instance from 1 to 5, or even from 1 to 3. In the instant example, n0is equal to 2. In figure 8, some low-length end branches of the graph are identified, and labelled as Br 1 - Br4, for the sake of illustration. For the end branch Br4, for instance, the leaf node is node 5, the common node is node 9, and the branch comprises of two nodes, namely node 5 and node 4.

[0091] Regarding the maximum length of low-length end-branch, n0, the ranges given above (from 1 to 5, or even from 1 to 3, for instance equal to 2) correspond to good compromises between :- avoiding excessive beam-path fragmentation or region fragmentation, caused by the presence of fringe hatch-lines, and- avoiding excessive graph condensation, which would lead to very big (and less numerous) regions, thus complicating (and possibly making non-optimal) the beam-to- region assignment.

[0092] When two or more nodes, corresponding to hatch-lines that are located in the extension of each other, are merged together, if all the nodes belong to low-length end branches (like nodes 1 , 2, 4 and 6, or like nodes 3, 5 and 7), they are all merged together and replaced by a single node, in the condensed-graph (see figure 9, where the series of nodes 1 , 2, 4 and 6 is replaced by a single node, and the same for the series of nodes 3, 5 and 7).

[0093] But if one of the nodes of the series may not belong to a low-length end branch (like node 9, for the set of nodes 8, 9 and 10), this node then belonging to the graph skeleton. In such a case, the other node(s) of the series (nodes 8 and 10 in this example) are merged, or in other words condensed, onto the node belonging to the graph skeleton.

[0094] If the series of nodes that may be condensed together comprises more than two graph skeleton nodes (each having, inter alia, a deepness lower or equal to n0), in addition to low- length end branches nodes, then, the low-length end branches nodes of the series arecondensed onto the graph skeleton node that is the closest, among the two graph skeleton nodes in question (closest in terms of distance between hatch-lines).

[0095] In this condensing process, the condensation may, like here, start with the nodes of the branches that are just before the leaf nodes, and then be continued with the leaf nodes.

[0096] Figure 9 shows a part of the condensed-graph Gc, obtained by condensing the graph G according to the rules presented above. As can be seen in this figure, many short branches, corresponding to fringe hatch-lines, have been condensed thanks to this procedure.

[0097] A particular way to achieve this condensation is now presented with reference to figures 10 to 14 and 11 B to 14B. In this exemplary case, the condensation is achieved first for low- length end branches that are located above the graph skeleton (figures 10 to 14), and then for the ones located below (figures 11 B to 14B).

[0098] First, it is noted that, as the hatch-lines all extend parallel to the hatching direction Dh, they can be sorted according to their position along the direction Ds which is perpendicular to Dh, that is according to a coordinate that identify the positioning the hatch-line considered along Ds. This coordinate is called the “height”, in the following. Two hatch-lines that have the same “height”, or level, are located in the extension of each other (i.e.: they belong to a same, virtually continued hatch-line). In this regard, it is noted that a given region may be extend from a first hatch-line (at a first level) to a last hatch line (at a last level), but that it may also extend from two or more first hatch-lines (if these first hatch-lines are at the same, first level) to two or more last hatch lines (if these last hatch-lines are at the same, last level of the region).

[0099] In the graph G, as represented in figures 10 to 14, the nodes are arranged so that their height in graph G (their vertical coordinate) is representative of the “height” of the corresponding hatch-lines (i.e.: representative of the coordinate locating the hatch-line, along Ds). The terms “above” and “below” refer to the height of the nodes relative to the skeleton of the graph.

[0100] Figure 10 shows the same part of the graph G as in figure 8. The nodes, that are grouped in one of the rectangles, in figure 10, have the same height (and so, they may potentially be condensed together). In figures 10 to 14, the nodes numbering is the same as in figure 8.

[0101] In figure 1 1 , the leaf nodes that are above the graph, called “maximal nodes”, are each surrounded by a circle. Each node, which is directly connected to a maximal node and which is not a fork node, is called a “premaximal node”. Premaximal nodes are surrounded by a dotted circle, in figure 1 1. In figure 12, common nodes, corresponding to the low-length end branches considered, are each identified by a square surrounding the node. The deepness of each node is defined as explained previously (distance to the closest common node).

[0102] The nodes are then condensed as explained above (but in a first time, only for nodes that are above the graph skeleton). The nodes that are condensed, together and / or ontograph skeleton nodes, are the maximal and premaximal nodes. The condensation is carried on first for the premaximal nodes (which corresponds to the passage from figure 12 to figure13) and then for the maximal nodes (which corresponds to the passage from figure 13 to figure14). The condensation onto a node that is not a maximal or premaximal node (like the condensation of node 14 onto node 15) is carried on only if said node (which is not maximal or premaximal) has a deepness lower than or equal to n0.

[0103] Then, the condensation is carried on for the low-length end branches that are located below the graph skeleton. In this particular embodiment, to achieve the condensation for such lower branches, the same rules as above are applied, but on the dual graph G* corresponding to G, instead of being carried on directly on G. Figure 11 B shows a part of G*. Figure 12B shows the maximal and premaximal nodes (maximal and premaximal on G*) that may be condensed. Figure 13B shows G* after condensing the premaximal nodes, and figure 14B shows the dual graph after having also condensed the maximal nodes (which corresponds to the dual of the finally condensed-graph Gc). step of graph segmentation s22

[0104] The graph segmentation is carried out on the condensed-graph Gc. It is carried on so that at least some of the regions defined, possibly most of them, or even all of them comprise of an ensemble of successive nodes of the graph, said ensemble of successive nodes not forming any fork, that is, forming a same branch that has just two ends.

[0105] This is illustrated in figure 16: each region of the graph Gc, R1 to R8, forms a single branch, with just two end nodes. Conversely, if the regions R3, R7 and R8 had been merged together to form a single region, for instance, this single region would have formed a fork (would have had a fork shape), with three ends, which is not desirable in terms of beam scanning efficiency of the region.

[0106] In the exemplary embodiment presented here, each region is delineated at each end, either by a leaf node LN of the condensed-graph Gc (in which case the leaf node is included in the region) or by a fork node FN of the condensed-graph (the fork node being not necessarily included in the region). For instance, region R1 is delineated on one end by a leaf node LN, and on the other end by a fork node FN (included in region R1 ). And region R3 is delineated on one end (upper end) by a fork node FN (not included in R3) and on the other, lower end by another fork node FN (included in R3).

[0107] Figure 17 shows the result of the step of partitioning s2, for the same portion of the section S than in figure 6, the different regions (i.e.: the different group of hatch-lines) visible in figure 17 being R1 , R2, R3, R4, R6, R7 and R8.

[0108] In alternative embodiments, the step of graph condensation could be omitted, the step of graph partitioning being still implemented (as above presented), but based on the graph G instead of begin based on the condensed-graph.Step s3

[0109] Once the ensemble E of hatch-lines is partitioned in regions, each beam is assigned a sequence of regions, this beam-to-region assignment being carried on so as to minimize the total idle time, in this exemplary embodiment.[001 10] To this end, a regions-graph Gr representing the ensemble of regions is first determined (figure 18). In this graph, each region defined in step s2 (for instance the regions R1 , R3, R4, R7,...) is represented by a node of the regions-graph Gr.[001 11 ] Regarding the edges of the regions-graph Gr, they are each attributed a length dij, i and j being the indexes of the two regions Ri and Rj linked by the edge in question, dij is representative of (for instance proportional, or even equal to) a distance between Ri and Rj. This distance is for instance the distance between the respective centers (e.g.: the respective centroids) of these two regions, or the distance between the respective upper vertices (or lower vertices) of these two regions, or another distance representative of the spacing (in the x,y plane) between these two sets of hatch-lines.[001 12] For the sake of clarity, just a few of the edges of the regions-graph Gr are represented in figure 18.[001 13] Once the regions-graph Gr determined, the minimization of the total idle time is carried on as a Travelling Salesman Problem based on the regions-graph Gr, problem in which the total distance travelled to visit all the nodes of graph Gr, exactly once, is minimized. This Travelling Salesman Problem (TSP) procedure is either a single agent or a multi-agents TSP, the number of agents being equal to the number N of beams of the device 1.1.[001 14] Here, this optimization is an optimization under constraint, taking into account the following constraint: for any couple of two of said regions for which one of the two regions is located downstream of the other region with respect to the gas flow F, the two regions are not processed by the beams at the same time. Avoiding simultaneous melting in two such regions is beneficial, as beam-melting usually creates smoke or particulates emission, that are evacuated by the gas flow. And so, it is desirable not to process a region located downstream of another region already being processed. Other constraints can be taken into account during this optimization procedure, instead of or in addition to the above-mentioned gas flow constraint. For instance, the optimization can be carried on such that, for two beams that are solidifying two zones of the powder layer simultaneously, the distance between the two zones being simultaneously solidified is above a given threshold (to prevent contamination by smoke or dust).[001 15] Different procedures are possible, for such a TSP optimization, for instance an Ant Colony Optimization procedure, a GRASP procedure, an evolutionary-based procedure such as a genetic algorithm, or even a so-called brute force optimization. Among these procedures, the inventors have observed that a genetic algorithm based procedure is well adapted, for thiskind of optimization problem (given the order of magnitude of the number of nodes and edges in Gr - typically between hundred and thousand or a few thousands the fact that it is a multiagent TSP when multiple beams are employed, and that it is an optimization with one constraint - related to the flow direction - or a few constraints but not many constraints).[001 16] As known in this technical field, a genetic algorithm starts with the generation of a population of several different graph-tours (initial parents). Two parents are then combined with crossover, and the resulting “child” graph tour may also experience mutation and is then given a fitness-score (representative of the total length travelled). On basis of the fitness-score, a part of the population is discarded and the rest is used to create a new generation.[001 17] For the examples presented below with reference to figures 19 to 24, the optimization algorithm used in step s3 is a genetic algorithm.[001 18] In alternative embodiments, the beam-to-region assignment data could be determined so as to directly minimize a total process time, instead of minimizing a total idle time. In such a case, a TSP optimization, based on the regions graph Gr, can also be employed, the nodes of the graph being then attributed a process time (or an equivalent distance, representing this process time) corresponding to the region process time for the region considered, for instance. In other alternative embodiments, in step s3, the beam-to-region assignment data could be determined so as to optimize a homogeneity of temperature (or of heat transfers) within the layer of material, or so as to optimize a criteria mixing a temperature homogeneity indicator and the total idle time or process time.[001 19] After step s3, the partition of the ensemble E of hatch-lines, into a plurality of distinct subsets (i.e.: into different regions), may be modified. The initial grouping of elementary segments as a single hatch-lines could also be modified. In particular, one or more regions (or even one or more hatch-lines) could each be subdivided, for instance into two new smaller regions (resp. into two smaller hatch-lines). Optionally, step s3 could be executed again after such a re-partitioning, to optimize again the idle time (based on this new partitioning ), the process time, or another characteristic of a process of manufacturing the section S of the part. Such a subdivision of one or more of the regions, into smaller ones, may be achieved when a significant unbalance between respective process times of different regions is detected. Indeed, subdividing a very large region into smaller ones helps finding a beam-to-region assignment for which the different beams operate evenly, which is favorable in terms of total idle time or total process time.Exam les

[0120] Figure 19 is a perspective view of stamping inserts built by powder-bed beam fusion manufacturing, using the device 1.1 , which comprises four laser beams. The total volume of the four stamping inserts is 837 cm3. They were built according to beam-to-region assignmentdata determined using the method presented above, which led to an average productivity of 55.5 cm3 / h (the manufacturing lasting 15 hours and 10 minutes).

[0121] Figure 20 is a top view of a section of the four stamping inserts. The regions, defined for this section, are represented in different shades of grey. A beam label B1 , B2, B3 or B4, is indicated for each region, specifying which beam was assigned to the region in question. The gas flow direction is identified by the arrow F.

[0122] Figure 21 is a top view of the same section as in figure 20, but with a partitioning into regions and a beam-to-region assignment achieved using a commercial software for additive manufacturing planning, instead of using the instant method. A manufacturing achieved using this commercial software, for identical manufacturing parameters (same powder, same hatchlines pitch, same scanning speed), except the partitioning and beam-assignment, lasted 19h, instead of 15h10min (almost 20% longer). And the inventors have observed that the total idle time was about 20% of the total process time (the lasers thus being clearly under-used), when using the commercial software. For the sake of comparison, with the method according the instant technology, the total idle time was below 5% of the total process time.

[0123] Figure 22 is a perspective view of an example of parts, to be manufactured using the device 1.1. These parts are nozzles intended to protect a measurement device.

[0124] Figure 23 is a top view of a section of one of these nozzles. The regions defined for this section are represented in different shades of grey. A beam label B1 , B2, B3 or B4, is indicated for each region, specifying which beam was assigned to the region in question.

[0125] Figure 24 is a top view of the same section as in figure 23, but with a partitioning into regions and a beam-to-region assignment achieved using the commercial software above mentioned, instead of using the method according to the instant technology. In practice, the total process time is 35% shorter when using the method according to the instant technology rather than using this commercial software. And the total idle time is smaller than about 5% of the total process time with this method, while it is of about one third of the total process time, for the commercial software segmentation / assignment.

Claims

1 . A method for determining beam-to-region assignment data for a powder-bed beam fusion device (1.1 ) suitable to emit one or more beams for solidifying selectively some zones of a powder layer, the method comprising the following steps, executed by a computer (1.6):- s1 ) Acquiring data specifying an ensemble (E) of hatch-lines (L) for the one or more beams (B1 -B4) to follow to gradually solidify zones of a powder layer (1.7), to build a section (S) of a part (P),- s2) partitioning the ensemble (E) of hatch-lines into a plurality of distinct subsets, each gathering some of the hatch-lines and forming a region (R1 - R8) to be processed by one of the one or more beams (B1 -B4),- s3) Determining the beam-to-region assignment data, which specify, for each of the one or more beams (B1 -B4), a sequence of regions to be processed by the beam considered, among the ensemble of said regions (R1 - R8), step s3 being achieved using an optimization procedure configured to optimize at least one characteristic of a process of manufacturing said section of the part using said one or more beams, and wherein step s2 comprises: o s20) building a graph (G) representing the ensemble (E) of hatch-lines, in which:• each hatch-line is represented by a node of the graph, and• for any couple of two hatch-lines that are laterally adjacent to each other, the two associated nodes of the graph are linked by an edge of the graph; partitioning the graph (G), or partitioning a condensed graph (Gc) derived from said graph (G), to define said regions, so that at least some of the regions each comprise of an ensemble of successive nodes of the graph (G), or of the condensed graph (Gc), said ensemble of successive nodes not forming any fork .

2. A method according to claim 1 wherein step s2 comprises, for at least some of the regions: (s22) grouping some of the hatch-lines (L) to form the region considered, said hatch-lines being each untruncated, and being laterally adjacent to each other, located one aside the other, one after the other successively.

3. A method according to claim 2 wherein, for at least some of the regions, the hatch-lines located one aside the other are distributed, one after the other successively, from a first hatch-line of the region to one or more last hatch-lines of the region, and wherein the first hatch-line, or one of the one or more last hatch-lines of the region, is one of:- a section-edge hatch-line (L3, L5, L7), that is a hatch-line which, on one side, has no laterally adjacent hatch-line;- a frontier hatch-line (Lf), that is a hatch-line having, on one side at least, more than one hatch-line (La, Lb) laterally adjacent to the frontier hatch-line (Lf).

4. A method according to claim 2 or 3, wherein step s2 comprises:(s21 ) grouping together, in one of said regions, at least two of the hatch-lines (L5, L7) that are located in the extension of each other, one after the other in a direction (Dh) parallel to the hatch-lines (L), provided that at least one of the two hatch-lines (L3, L5) has, on one side, no laterally adjacent hatch-line.

5. A method according to anyone of the preceding claims, wherein step s2 comprises:- a step of graph-condensation s21 ), during which the condensed-graph (Gc) is determined from said graph (G), the step of graph-condensation comprising: o merging one of the nodes (5) of the graph on another node (7) of the graph, when the two nodes (5, 7) are associated to two hatch-lines (L5, L7) that are located in the extension of each other, one after the other in the direction parallel to the hatch-lines (Dh), when at least one of the two nodes (5, 7) is a leaf node of the graph,- s22) partitioning the condensed-graph (Gc) to define said regions, so that at least some of the regions each comprises of an ensemble of successive nodes of the condensed- graph, said ensemble of successive nodes does not forming any fork.

6. A method according to claim 5, wherein the step of graph-condensation s21 ) comprises: identifying low-length end branches ( Br 1 - Br4) in the graph (G), o an end branch of the graph being a part of the graph extending from a leaf node (3) of the graph to a common node (9) associated to that leaf node, said common node (9) being the first fork node encountered on the graph when starting from the leaf node considered (3) and following the graph, the end branch not including said common node, o a low-length end branch (Br1 - Br4) being an end-branch whose total number of nodes is below or equal to a given maximum number of nodes n0, for each low-length end branch identified, for each node (3) of the low-length end branch considered, merging said node (3) onto another node (5) of the graph if thesetwo nodes are associated to hatch-lines (L3, L5) that are located in the extension of each other, one after the other in the direction parallel to the hatch-lines (Dh).

7. A method according to claim 6, wherein n0is from 1 to 5, or even from 1 to 3.

8. A method according to anyone of claims 1 to 4, wherein said regions are defined by partitioning said graph (G) so that the ensemble of successive nodes forming each of said regions has only two ends and is delineated at each end either by a leaf node or by a fork node.

9. A method according to any of claims 5 to 7, wherein the condensed-graph (Gc) is partitioned so that the ensemble of successive nodes forming each of the regions (R1- R8) has only two ends, and is delineated at each end, either by a leaf node (LN) of the condensed-graph or by a fork node (FN) of the condensed-graph.

10. A method according to any of the previous claims,- further comprising acquiring a direction of a gas flow (F), to be blown over the powder layer (1 .7) during the beam melting of zones of the powder layer,- and wherein, in step s3, said optimization procedure is configured to minimize said characteristic of the process of manufacturing said section of the part, while taking into account the following constraint: for any couple of two of said regions, for which one of the two regions is located downstream of the other region with respect to the gas flow (F), the two regions are not processed by the beams at the same time.1 1. A method according to any of the previous claims, wherein the characteristic of the process of manufacturing said section of the part, which is optimized in step s3 is:- a total idle time, representative of the sum of the jump times for the one or more beams, each jump time being a time required, for one of the one or more beams, to go from one of the regions belonging to the sequence of regions to be processed by that beam, to the next region in said sequence;- or a total process time for building said section.

12. A method according to claim 11 , wherein step s3 comprises:- building a regions-graph (Gr) representing the ensemble of regions (R1 - R8) determined in step s2, each of said regions being represented by a node of the regions- graph (Gr), each edge of the regions-graph, that link two of the nodes of the regions-graph, being attributed a length (dy) representative of a distance between the regions corresponding to the two nodes considered,- and wherein the optimization procedure of step s3 is a single-agent or multi-agents Travelling Salesman Problem optimization procedure, based on said regions-graph (Gr), and configured for minimizing a total distance travelled on said regions-graph for visiting all the nodes exactly once.

13. A method according to claim 12, wherein the optimization procedure is a Genetic Algorithm optimization procedure.

14. Method for manufacturing a part (P) using a powder-bed beam fusion device (1.1 ), the method comprising:- for one or more sections of part (P), determining beam-to-region assignment data associated to the section (S) considered, according to the method of any of the previous claims,- s4) building said one or more sections, according to the beam-to-region assignment data.

15. Calculator (1 .6) comprising at least a processor and a memory, configured, for instance programmed to execute the method according to any of claims 1 to 13.

16. Powder-bed beam fusion device (1.1 ), comprising:- a calculator (1 .6) according to claim 15,- a support (1.2), for layers of raw material powder (1.9, 1.8, 1.7) to be successively deposited to produce a part (P),- a beam source (1.4) suitable to emit one or more beams (B1 - B4) with controllable emission directions, for totally or partially melting powder layers,- a controller (1 .5) operatively connected to the beam source (1 .4) and to the calculator (1.6) and configured for controlling the beam source (1.4) according to the beam-to- region assignment data received from the calculator (1 .6).

17. Computer program comprising instructions whose execution on a computer make the computer to execute the method according to any of claims 1 to 13.

Images