BE1027711B1 - TURBOMACHINE COMPRESSOR STAGE - Google Patents

Abstract

A method of designing a low pressure compressor compression stage (34) comprising (4), the stage (34) comprising: a rotor support (26), from which rotor blades extend. (24) axially followed by a stator support (32) from which extend stator vanes (25). The connecting surface between two circumferentially adjacent blades (24, 25) of the support (26) or of the support (32) comprises irregularities, also called “3D contouring”. The design process includes parameterizing these bonding surfaces using splines and simulating compressor efficiency as the parameters vary.

Description

, Ç BE2019 / 5739 Description

TURBOMACHINE COMPRESSOR STAGE Technical field The invention relates to an irregular geometry in an inter-blade passage of a turbomachine compressor and more particularly to the design method of this geometry. The invention relates in particular to the design of a turbomachine compressor stage, such as a turbojet or an aircraft turboprop. PRIOR ART Patent application EP 3 477 050 A1 discloses a turbojet with a compressor comprising a rotating ferrule or a fixed ferrule which is provided with a hump and a hollow in each inter-blade passage. These irregularities on the air guiding surface - also called “3D contouring” - make it possible to improve the flow of the flow and in particular to limit the separation of the air flow in the vicinity of the upper surface and / or the air flow. blade roots. In this known design, the boss and the hollow are axially aligned and positioned in an upstream half of the vanes. In this document only a single row of vanes (stator or rotor) is considered and the emphasis is on the design of the inter-vane space. However, if this design improves the air flow at the root of the blades as it passes between these blades, it does not necessarily improve the flow upstream or downstream of this row of blades. The design of the geometry is carried out locally and "manually": the designer imagines a rotor or stator geometry and then experimental tests follow to validate the geometry. The process is tedious and introduces a human bias that limits the optimization of the geometry by the human imagination. || There is therefore room for improvement in the flow of the air flow in the compressor. Summary of the invention Technical problem

> BE2019 / 5739 The object of the invention is to propose a design method implemented by computer which minimizes the material resources necessary to obtain a geometry optimizing the flow at the root of the blade in a compressor (flow rate, pressure, separation, compression ratio, pumping margin).

Technical solution The subject of the invention is a method for dimensioning, implemented by computer, of the connecting surfaces of a compressor stage, the stage comprising: a rotor support and an annular row of rotor blades extending radially outwardly from the compressor stage. rotor support, the rotor support defining a rotor connection surface between two circumferentially adjacent rotor blades; a stator support and an annular row of vanes extending radially outwardly from the stator support, the stator support defining a stator connection surface between two circumferentially adjacent stator vanes; the annular row of stator vanes being positioned directly downstream of the annular row of rotor vanes; the dimensioning method comprising an initialization phase comprising: the definition of parameters, preferably in computer-assisted drawing software, CAD, defining a plurality of splines circumferentially distributed on the rotor connection surface; the definition of parameters, preferably in the CAD software, defining a plurality of splines circumferentially distributed on the stator connection surface; determining a mesh of the geometry of the connecting surfaces and possibly of the respective blades; the determination of the experimental design comprising intervals of authorized values for the parameters, sampling laws in these intervals, grid criteria and possibly a criterion for interrupting the simulation; the method then comprising a calculation phase with the following steps: exporting the mesh to fluid mechanics calculation software, CFD; simulating the air flow in the compressor stage and calculating the efficiency of the compressor stage; and as long as a possible interruption criterion is not reached, the variation of the values of one or more parameters in accordance with the sampling law, the definition of a corresponding mesh and the iteration of the calculation; the method then comprising a phase of choosing one or more geometries of connecting surfaces as a function of the results obtained during the calculation phase. The rotor support may be a ferrule, a drum, an individual blade attachment platform, or a portion of a disc or bladed wheel. The stator support may be a ferrule or an angular segment of a ferrule, or an individual blade attachment platform. By “circumferentially adjacent” is meant two blades of the same annular row of blades which are directly adjacent circumferentially.

The geometry of the connecting surfaces can be repeated step by step between all pairs of circumferentially adjacent blades.

The connecting surfaces can each be provided with axisymmetric portions and non-axisymmetric portions, the latter comprising at least one bump and / or at least one hollow.

The splines here are class C2 curves of space, piecewise polynomial. The choice of splines makes it possible to avoid excessively angular curves which would require a mesh that is too fine or inappropriate for the simulation. Among all the curves allowing modeling, the present application uses a specific choice with C2 splines.

It is important to stress that none of the design steps are obvious or result from an arbitrary choice. The knowledge and experience of the aeronautical engineer were necessary to develop the present invention for which the person skilled in the art is indeed an aeronautical engineer: the relevance of the choice of the boundary conditions (pressure, input speed), the choice of the objectives to be achieved, the constraints imposed, the choice of criteria allowing the creation of a robust mesh, the choice of the field to be analyzed (experimental design), the choice of sampling and interpolation algorithms, and up to '' checking the consistency of the results.

According to advantageous modes of the invention, the method can comprise one or more of the following characteristics, taken in isolation or according to all the possible technical combinations:

the parameters defining the splines of the rotor support comprise the coordinates of two respective points through which a given spline passes and the parameters defining the splines of the stator support include the coordinates of two respective points through which a given spline passes.

The choice of class C2 splines with predefined boundary conditions allows, using only two points, to define the equations of each spline in space.

Thus, the set of points of the spline and therefore a curve of the bonding surface can be defined by only six values (the three coordinates of each of the two points). Little memory capacity is therefore necessary to define a curve and define its variations.

This is a good compromise between the necessary computation time and the expected precision of the geometry; - in a given inter-blade space, the splines are circumferentially spaced and are three in number for the rotor support and three for the stator shell.

They can pass through two respective fixed points in addition to the two parameterized points which are caused to vary.

This allows for greater versatility in the shapes that the splines can take without requiring more computing capacity; - the values of the authorized parameters are such that the splines do not move away radially from an axisymmetric portion of the connecting surface by more than 8% of the length of the chord of the respective blades.

This limitation prevents the simulation from unnecessarily exploring technically irrelevant areas.

This limit is high enough to leave flexibility and autonomy to the simulation while limiting the simulation to technically feasible solutions.

Alternatively, this limit can take any value between 4% and 10%; - the mesh criteria include limits imposed on the characteristics of the mesh, such as orthogonality, mesh angles, expansion rate.

In the present application, orthogonality is understood to mean the local property of a mesh whose segment connecting two centers of gravity of two adjacent meshes is perpendicular to the edge that these two meshes have in common.

The mesh angle is the angle formed by two edges of the same mesh.

It can be limited to a minimum and / or maximum value.

The expansion rate is the tolerable variation (mini or maximum) between two successive meshes (in time during the calculation) or between two adjacent meshes of the same mesh (in space). It is in fact important to criticize the mesh so that it remains robust, that is to say that it changes little when the geometry changes little and the mesh adapts well when the geometry changes a lot so as not to to present deformed meshes (so that the results obtained by computation by finite elements are close to reality). A stochastic mesh would require computing resources and more time to be defined.

The meshes can be calculated step by step or be based on a known mesh for a neighboring geometry.

In the terminology used here, the mesh criteria are the conditions imposed on all the meshes during a calculation phase.

This terminology is to be distinguished from the characteristics which qualify a given mesh (types of meshs, positions of nodes, etc.) during a step of creating the mesh (see step 204 below); - alternatively to the imposition of limits for the mesh criteria, the method can include a step of monitoring the characteristics of the mesh and when the characteristics of the mesh deviate from admissible values, the simulation is not carried out for the sample parameters or the results obtained by the simulation are annotated with corresponding information.

Thus, instead of forcing a mesh with certain limits, the mesh is the prerogative of the computer according to basic mesh constraints (type of mesh, mesh refinement, etc.). Only a subsequent check is carried out, thus limiting the necessary computational resources.

Also, the fact of discarding inconsistent results resulting from a non-conforming mesh allows rapid convergence of the results towards optimums, in particular if the results are discarded during or before the calculation; - after the calculation phase, a response surface is built by interpolation of the discrete results from each of the iterations, representing the efficiency calculated during the calculation phase as a function of the parameters and the step of choosing one or more geometries binding surfaces includes determining and selecting one or more sets of parameters for which the response surface has a local or global extremum.

The choice will be made for example by means of an optimization algorithm because once the response surface has been constructed, it is possible to know the entire space of results and therefore the search for the optimum is easy.

The use of a response surface avoids the trap of a genetic algorithm which could converge towards values with local optima while ignoring better results further away in space.

In a method not detailed here of direct convergence, an interruption criterion can be defined and is reached when the variation of the yield calculated between two successive iterations is less than a predefined threshold.

This makes it possible to avoid too long calculation times if the optimums are identified from the start of the calculation; In a space of dimension N + 1 where N is the number of parameters defining a given geometry, the response surface is a surface of dimension N and the N + 1st value is the calculated efficiency of the compressor.

The iterations give discrete results of efficiency and the response surface is an interpolation, for example in the sense of least squares or by kriging (minimal variance) of these discrete results; the succession of the initialization, calculation and choice phases defines a first calculation session, the method continuing with a second calculation session comprising a second initialization phase, a second calculation phase and a second choice phase , the experimental design of the second calculation session being different from the experimental design of the first calculation session.

This second calculation session makes it possible to refine the results of the first calculation session.

Thus, it comprises intervals of authorized values for the parameters and / or sampling laws of the parameters and / or of the mesh criteria and / or a criterion for interrupting the simulation which are different from what is defined in the experimental design of the first calculation session.

Indeed, too fine parameter variations or too fine sampling from the first computation session could prove to be needlessly consuming computation time or resources.

Also, the fixed points of the splines can be changed from session to session;

- the interruption criterion is a predefined iteration number or a predefined calculation time; the simulation step further comprises the calculation of the total pressure and / or of the flow rate at the roots of the blades and the samples giving a result lower than a predefined threshold of total pressure and / or flow rate are rejected.

As the aim of the invention is to determine a surface which is used in a system (the compressor), it is necessary that this surface can perform a physical function which meets the specifications.

A necessary condition for the simulation to be able to end is that at least one technical solution meeting minimum values of total pressure and flow can be delivered.

This condition is not sufficient because the first technical solution found chronologically during the simulation is not necessarily the best.

Simulations that do not meet these criteria can be erased from memory to limit hardware resource requirements.

The total pressure and the flow rate can also be combined (linear combination for example) with the efficiency to optimize an objective which is the result of these three physical values; - the sampling law is of the Latin hypercube type.

This technique allows credible computation results, in particular for a large number of variables, as is the case here, and enables the working domain to be correctly represented by a minimum of computations.

Alternatively, other methods can be used, such as that of factorial designs.

The invention may also relate to a method for producing a rotor or a stator in accordance with the geometry adopted during the method according to one of the preceding embodiments.

The invention also relates to a compressor rotor for a turbomachine, comprising a rotor support and an annular row of rotor blades extending radially outwardly from the rotor support, the rotor support defining a rotor connection surface between two circumferentially adjacent rotor blades, the rotor connection surface conforming to a rotor support connection surface geometry directly obtained by the sizing method of one of the embodiments described above.

The invention also relates to a compressor rectifier for a turbomachine, comprising a stator support and an annular row of stator vanes extending radially outwardly to the stator support, the stator support defining a stator connection surface between two circumferentially adjacent stator vanes, the stator connection surface conforming to a stator support connection surface geometry directly obtained by the sizing method of one of the embodiments described above.

The invention also relates to an example of a rectifier obtained by the method and which can be characterized as follows.

Compressor rectifier for a turbomachine, comprising a stator support and an annular row of stator vanes extending radially outside the stator support, the stator support defining a stator connection surface between two circumferentially adjacent stator vanes, the connection surface being provided with axisymmetric portions and non-axisymmetric portions, the latter comprising a hump and a hollow between two circumferentially adjacent vanes, characterized in that the hump of the stator support is contiguous to the intrados of a stator vane and extends axially over the less 75%, preferably over 100%, of the chord of the stator vanes and the hollow of the stator support is disposed circumferentially at approximately equidistant from two circumferentially adjacent vanes.

According to an advantageous embodiment, the top of this bump is attached to or close to the lower surface of the stator vane.

According to an advantageous embodiment, the hollow is arranged axially at a distance from the leading edges of the stator vanes of approximately 75% of the chord of the stator vanes.

Finally, the invention relates to a compressor stage for a turbomachine, comprising a rotor and a rectifier as described in the two preceding paragraphs, the stator vanes of the rectifier being arranged axially directly downstream of the rotor vanes of the rotor.

Advantages The method according to the invention makes it possible to minimize the computer resource requirements (in memory or processor capacity, and in computing time) in order to obtain physically usable results.

The association of non-axisymmetric irregularities on the rotor and on the stator directly downstream of the rotor allows a synergistic effect of improving the flow of the flux.

Precise choices of bumps or hollows on the rotor, combined with precise choices of bumps or hollows on the stator make it possible, on the one hand, to increase the pressure gradient across the passage between rotor blades, thus inducing a more strong tangential deviation of the flow (tangential component of the speed of the flow at the outlet of the rotor), from the lower surface of a blade towards the upper surface of its neighbor, and on the other hand, to attenuate the pressure gradient in the inter-blade passage of the stator, by accelerating the flow on the intrados and by decelerating it on the extrados.

Thus, the secondary flow is reduced and the axial realignment of the flow at the exit of the stator is facilitated.

The combination of the bumps and valleys of the rotor and stator impact the flow, better than would be enhanced by independently designed bumps and valleys on the rotor or stator.

Thus, not only is the separation of the flux at the root of the rotor blades is reduced, but also the flux is "prepared" to travel through the stator.

On the stator side, the irregularity attenuates the pressure gradient across the inter-blade passage by accelerating the flow on the lower surface and decelerating it on the upper surface.

This facilitates the axial realignment at the output of the rotor + stator stage.

Also, the irregularity of the stator does not “slow down” the air flow leaving the rotor.

There is therefore a doubly positive effect of the two irregularities acting together.

The dimensioning method according to the invention makes it possible to overcome the technical difficulties linked to the calculation time and to the need for capacity of the computer means used.

Indeed, on one side of the prior art spectrum, the simulation is done "manually" for a geometry given by the engineer for the rotor.

After calculation, the engineer instinctively modifies the geometry and a new simulation is launched.

Once the rotor has been optimized, the operation is repeated for the stator.

This technique minimizes IT resource requirements but maximizes the time required (or equivalent sizing time minimizes the number of samples tested). On the other side of the spectrum of the state of the art, the simulation is done in "brute force": any type of geometry is tested without distinction and without limit on the shape of the bonding surface.

The simulation can last several months without convincing results.

Without guidelines for the parameters, the simulation can “lock itself in” to a local optimum without investigating other values.

By comparison, the present invention offers the best compromise between calculation time and relevance of the results thanks in particular to the parameterization by splines and to the calculation on the entire stage.

Also, the choice of the techniques employed in the dimensioning method according to the invention makes it possible to overcome these difficulties in terms of time and resources.

In particular, the iterations and the analysis of a response surface rather than a classic genetic algorithm makes it possible to reduce the calculation time because it is not necessary to wait for the result of a calculation and to process it in comparison. other results, to adjust the settings accordingly.

The use of three splines parameterized by the position of two points makes it possible to be limited to a space of 36 dimensions.

Each additional dimension multiplies the computation time exponentially, and the computation according to the invention comprises several million iterations in a 36-dimensional space and is therefore industrially realistic with a computation time of between a few hours and a few days.

Also, imposing criteria on the mesh makes it possible to make it robust and to avoid the risks of unusable calculations.

The choices of the techniques used in the process are specifically adapted to the aeronautical field and in particular to the calculation of a compressor stage.

Brief description of the drawings

Figure 1 shows a diagram of a turbomachine compressor. Figure 2 shows the compressor stage seen in the radial direction. FIG. 3 describes a method according to the invention. Figures 4 and 5 illustrate the setup. Figures 6 and 7 show two examples of geometries according to the invention, resulting from the method according to the invention. Description of the Embodiments In the description which follows, the terms “internal” and “external” refer to a positioning relative to the axis of rotation of an axial turbomachine. The axial direction corresponds to the direction along the axis of rotation of the turbomachine, the lengths being measured axially. Widths are measured by circumference. The radial direction is perpendicular to the axis of rotation. Upstream and downstream refer to the main flow direction of the flow in the turbomachine. The dimensions of the figures are not to scale and in particular the thicknesses or the radial dimensions are exaggerated to facilitate reading of the figures. Figure 1 is a sectional view of a compressor 4 of an axial turbomachine. A rotor 12 is coupled to a blower 16 and generates an air flow which divides into a primary flow 18 and a secondary flow 20 at a separation nozzle.

22. The secondary flow is accelerated to generate a thrust reaction useful in the flight of an aircraft. The rotation of the rotor around its axis of rotation 14 makes it possible to generate an air flow and to gradually compress the primary flow 18 up to the entrance of the combustion chamber (not shown). The rotor 12 comprises several rows of rotor blades 24, in this case three. The compressor 4 comprises several rectifiers, in this case four, which each contain a row of stator vanes 25. The rectifiers are associated with the fan 16 or with a row of rotor vanes 24 to straighten the air flow, so as to convert the speed of the flow under pressure, in particular under pressure - static.

The rotor vanes 24 can extend radially from a rotor support 26 which can be a dovetail platform, a bladed one-piece drum internal ring, or any other type of composite rotor support. The support can be an angular section describing less than 360 ° around the axis

14. The stator vanes 25 extend essentially radially from an outer casing 28. They can be fixed there and immobilized by means of fixing pins 30. They pass radially through the primary flow 18. Each row can comprise at least fifty vanes 24, 25.

Stator supports 32 are suspended from the internal ends of the stator vanes 25. The stator supports 32 can cooperate in a sealed manner with the rotor 12.

The compressor comprises three compression stages 34, each composed of a row of rotor blades 24 followed by a row of stator vanes 25.

FIG. 2 outlines two rotor blades 24A, 24B representative of an annular row of rotor blades and three stator vanes 25A, 25B, 25C representative of an annular row of stator vanes, seen radially from the outside.

The axis of rotation 14 is drawn at a figurative position to provide a spatial landmark.

The rotor blades 24 rotate in the direction indicated by the arrow 36. The general direction of flow of the flow is indicated by the arrow 18. This direction is given as a reference for this figure because locally and in particular between the rotor blades 24 and the stator vanes 25 the direction of flow differs from the direction 18.

The rotor vanes 24 are carried by the support 26 and the stator vanes 25 are carried by the stator support 32. The vanes 24, 25 respectively have a leading edge 40, 41, a trailing edge 42, 43, an extrados 44 , 45 and a lower surface 46, 47.

The support 26 and the support 32 define a respective connecting surface 48, 49, which guides the air flow radially internally and which extends from the upper surface

44, 45 of a vane 24A, 25A at the lower surface 46, 47 of a directly circumferentially adjacent vane 24B, 25B. This connecting surface 48, 49 comprises an axisymmetric portion 50, 51 (cylindrical, conical, ellipsoid, spherical, etc.) around the axis 14 and a non-axisymmetric portion 52, 53.

The non-axisymmetric portion 52 of the rotor support 26 may comprise one or more bump (s) and / or one or more hollows. The non-axisymmetric portion 53 of the support 32 may comprise one or more bump (s) and / or one or more hollows. These non-axisymmetric surfaces 52, 53 are illustrated schematically in Figure 2.

The connecting surface 48, 49, the upper surface of a vane 24A, 25A and the lower surface of a directly circumferentially adjacent vane 24B, 25B, define a respective inter-vane passage.

The blade chord, which is identical for all the blades of a given row of blades, is denoted CR (rotor chord) and CS (stator chord). The chord is a segment that runs from the leading edge to the trailing edge.

FIG. 3 illustrates a method according to the invention which can be followed to determine the geometry of the stator and of the rotor in an optimum manner.

The method begins with an initialization phase 100 during which initial parameters are chosen, in particular defining a “starting” geometry for the connecting surface and surrounding the variations made on this geometry to result in an optimized geometry. The geometry is parameterized as shown below using splines. An experimental design delimits the set of possible samples which are simulated, for example according to a Latin hypercube sampling. Constraints on the robustness of the mesh are applied. A mesh is generated for the starting geometry. The know-how and the experience of the aeronautical engineer are necessary to anticipate the results and problems and to prepare an adequate network. The mesh must be refined in certain places, certain types of mesh are more advantageous than others.

The method continues with a calculation phase 200. This phase iteratively comprises steps of calculating and updating the geometry in accordance with the experimental design. Iterations continue until an interrupt criterion is met. The calculation step 201 essentially comprises the simulation of the air flow in a turbomachine whose compressor stage conforms to the geometry of the iteration in question and the calculation of the efficiency of the compressor. In addition, the total pressure at certain points of the blade roots and the flow rate can be calculated. As long as the interruption criterion 202 is not reached, the geometry is modified 203, in accordance with the experimental design and a new mesh is generated 204. The interruption criterion can be defined in the initialization phase 100. I! can be a number of cycles, a minimum calculation result (efficiency, pressure, flow), a convergence of the results of several successive cycles (successive in the direction with neighboring parameters, not necessarily successive in the chronology of the cycles). The operator can in all cases intervene and carry out the end of the cycles. The mesh 204 is constructed in accordance with the rules established in step 100.

Alternatively, it can be built autonomously - although resulting from the starting mesh and the conditions imposed on the mesh - then tested in a step 205 against the rules defined during step 100. If the mesh does not conform, then computation is not carried out in the following cycle or the results are ignored or tagged as corresponding to a non-conforming mesh.

After the calculation step 201, there may be a post-processing step 206 during which the non-essential information following the process is deleted from the memory. Post-processing 206 could possibly include the step-by-step construction of a response surface.

Alternatively, a step of constructing a response surface 301 can be part of the third phase 300 of the process which consists of a choice of the optimal geometry.

The selection of optimal geometries can be done automatically from the response surface (for example by canceling partial derivatives). It can be done by the operator after an automatic pre-selection.

The method can implement several different software, CAD, CFD, calculation, finite element modeling, etc. The method therefore also comprises steps of exporting / importing 100 ’, 203’, 204 ’from one software to another software.

The method can comprise a step 400 of manufacturing a compressor stage whose connecting surfaces correspond to the geometry chosen in phase 300. The method can comprise a single session of phases 100, 200 and 300. Alternatively, the method can comprise several sessions, including a results refinement session.

Thus, after a first calculation session, a first selection of interesting geometry (and therefore of parameters), a new session can be carried out (see loop 302 ′ in FIG. 3) in which the initialization phase 100 is different from the initialization phase of the first session.

We could for example have a different starting geometry to see if the optimal geometry proposed is identical.

Sampling (variations in geometry between cycles) to be finer in areas close to the expected optima, and coarser in less interesting areas.

Also, different mesh characteristics are possible because knowledge is acquired during the first session allowing the user to anticipate the evolution of the mesh according to the evolution of the parameters and therefore to choose in an optimal way the mesh of departure to avoid peak or trough phenomena.

In particular, the results discarded from the calculation during the previous session can alert the user to the need to modify the starting mesh.

FIG. 4 repeats the illustration of FIG. 2, without the non-axisymmetric portions for greater clarity to illustrate the parameterization of the connection surfaces.

Parameterization can be performed using points 70-81 defining splines 82-87 contained in the surface of the bumps and / or valleys.

The points 70-81, for example 6 in number on the rotor support 26 and 6 in number on the stator shell 32, are positioned in pairs 70-72, 74-76, 78-80, 71-73, 75- 77, 79-81. In the “starting geometry”, these points can for example be “aligned” according to the camber of the respective blades 24, 25. In other words, seen in the plane of FIG. 4 - which is a view along a radius of the turbomachine - a curve parallel to the camber can connect the points two by two. In the example shown, the spline 84 is arbitrarily drawn equidistant from the splines 82 and 86. Alternatively, it can be 5, 10, or 20% of the inter-vane distance on the side of the spline 82 or the spline 86 Four 70-73, 78-81 of the respective six points 70-81 can be positioned in the middle surface of the vanes. The last two 74-77 of these six respective points are arranged equidistant from the two blades 24A, 24B, 25A, 25B. Each point is characterized by its coordinates in a cylindrical frame (axial position with respect to the leading edge 40, distance to radius 14, and circumferential position with respect to a reference point - for example the leading edge of the blade 24A).

The connecting surface 48, 49 passes through the splines 84 and 85 and its theoretical extension within the blades passes through the splines 82, 83, 86, 87. The splines 82-87 extend in the three directions of space. . Class C2 splines pass through two points of a pair of points.

By varying the coordinates of points 70-81, the splines vary and the geometry of the bond surface varies. 3x12 = 36 parameters are therefore sufficient to completely define the connection surface.

The splines can optionally pass through 12 additional points A to L. These points A to L are defined in the initialization phase 100 and their position remains unchanged during variations in the geometry. Thus, a more complex geometry can be studied without altering the requirements in computing resources. FIG. 5 illustrates for example the curve 84 projected in the plane denoted X: X in FIG. 4, in a flattened view of the curve 84. It shows in particular the points 74 and 76 and the tangents T74 and T76 to the curve 84. in this plane.

By varying the radial positions R74 and R76 of the points 74 and 76 and / or their circumferential positions, the tangencies T74 and T76 (in the X: X plane) vary automatically so that the spline continues to pass through the fixed points B and E .

Alternatively, one can vary the two coordinates (abscissa and ordinate in the X: X plane) of the tangency vectors T74 and T76 or impose them as fixed.

The same goes for the other curves 82-87. Thus, by varying the 36 parameters linked to the points, and by simulating the consequences on the air flow, it is possible to obtain a series of vectors with 37 dimensions, of which 36 are the parameters and the 37 ° "° is the calculated efficiency. .

Alternatively, the 37th element of the vector can be an efficiency weighting with other objectives to be optimized: total pressure, flow rate, pumping margin, etc.

These vectors allow the construction of the response surface (surface of dimension 36 in a space of 37 dimensions).

Two examples of geometries that can be obtained with the method according to the invention are summarized in the following figures. These two examples can be obtained with different initial conditions (initialization phase 300, choice of the starting geometry or of the constraints or objectives to be achieved) or can be two distinct optima resulting from the same response surface.

FIG. 6 shows in particular the level lines which mark the radial variations of the bosses 54, 56 and the hollows 55, 57 with respect to the axisymmetric portion 50, 51. For greater clarity, only an inter-blade space is shown with these bumps and hollows but the patterns are repeated regularly circumferentially around the axis 14, for all or part of the inter-blade spaces.

On the support 26, the bump 54 and the hollow 55 are located on either side of a line connecting the leading edge 40 of the vane 24B to the trailing edge 42 of the vane 24A. This line can substantially form an axis of antisymmetry for the bump 54 and the hollow 55.

The bump 54 has a top 60 materializing the point of the bump 54 furthest radially from the axisymmetric surface 50. The hollow 55 has a bottom 61 materializing the point of the hollow 55 furthest radially from the axisymmetric surface 50.

It is the same for the stator support 32: the bump 56 has a top 62 materializing the point of the bump 56 furthest radially from the axisymmetric surface 51. The hollow 57 has a bottom 63 materializing the point of the hollow 57 furthest away. radially of the axisymmetric surface 51. In the example of FIG. 6, the bump 54 extends circumferentially substantially from the leading edge 40 of the vane 24A to the leading edge of the vane 24B - that is, that is to say that it occupies at least 80% or at least 90% of the inter-vane distance (denoted T in FIG. 7) in the vicinity of the leading edges 40. The bump 54 also extends over the major part of the upper surface 44 of the blade 24A, that is to say more than 50% of the upper surface 44, and preferably more than 90% of the upper surface 44. It is contiguous to the upper surface 44 of the 'blade 24A and the top 60 may be in the vicinity of, or even in contact with, the upper surface 44.

Hollow 55 extends circumferentially over about 75% of the inter-vane distance and extends along the underside 46 of vane 24B over approximately the downstream half of vane 24B. It is attached to the intrados 46 of the blade 24B and the bottom 61 may be in the vicinity, or even in contact with the intrados 46.

On the stator support side 32, the bump 56 extends substantially from one leading edge 41 to the other and axially along the lower surface 47 of the vane 25A. The bump 56 may extend around the trailing edge 43 and move upstream on the upper surface for about one-fifth to one-third of the chord. The top 62 is attached to or close to the intrados 45, in particular at approximately 5% of the inter-blade distance (between 3% and 12%).

The hollow 57 is located approximately in the middle of the inter-blade space, that is to say equidistant from the blades 25A and 25B. It can extend between 40 and 60% of this inter-vane distance and be axially centered on an area approximately 75% (that is to say any value between 60 and 80%) of the length of the chord. , measured from the axial position of the leading edge. The height of the bump 56, defined by the distance from the top 62 to the axisymmetric portion 51 is between 40% and 60% of the height of the bump 54.

The bump 56 can rise 1.55 mm (+/- 20%) above the axisymmetric surface. The dip 57 can go down to -1.0mm +/- 20% below the axisymmetric surface.

FIG. 7 illustrates a second example of geometry obtained by the method according to the invention.

The elements identical to the first embodiment retain their reference numbers.

The separate elements are incremented by 100 by comparison with FIG. 6. In this example, the bump 154, which extends substantially from one leading edge to the other, has the top 160 and the hollow 155 has the bottom 161. In this example, the stator support 32 only comprises a bump 156 and no hollow.

The ratio of the depth of the hollow 155 to the height of the bump 156 is here approximately 55%. The T notation represents the inter-vane distance.

In this example, the hollow circumferentially occupies at most 25% of this distance and the bump occupies at least half of it.

The hollow extends in the downstream half of the stator support.

A theoretical line materializes a corridor between the bump 154 and the hollow 155, the corridor possibly being an axisymmetric portion.

Another line materializes the downstream termination of the bump and the hollow.

The two lines can be substantially perpendicular.

The examples illustrated here show a bump and / or a hollow, but the invention is not limited thereto.

Indeed, those skilled in the art will understand that the number of parametric points and the number of bumps or hollows may be different to result in more complex surfaces.

The same principles as those set out above also apply.

In the present application, the terms “hollow” and “bump” are used for convenience to qualify any geometry respectively going radially below or beyond a nominal geometry.

These terms are therefore synonymous with "depression" and "elevation". The examples of geometries illustrated here are described in more detail in application BE 2019/5738.

Claims (16)

Claims 1. Method of dimensioning implemented by computer, of connecting surfaces (48, 49) of a compressor stage (34, 134), the stage (34, 134) comprising: - A rotor support (26) and a row annular rotor blades (24) extending radially outwardly from the rotor support (26), the rotor support (26) defining a rotor connection surface (48) between two circumferentially adjacent rotor blades (24A, 24B); - A stator support (32) and an annular row of vanes (25) extending radially outwardly from the stator support (32), the stator support (32) defining a stator connection surface (49) between two stator vanes ( 25A, 25B) circumferentially adjacent; the annular row of stator vanes (25) being positioned directly downstream of the annular row of rotor vanes (24); the dimensioning method comprising an initialization phase (100) comprising: - the definition of parameters, preferably in computer-assisted drawing software, CAD, defining a plurality of splines (82, 84, 86) distributed circumferentially on the surface rotor link (48); - the definition of parameters, preferably in the CAD software defining a plurality of splines (83, 85, 87) circumferentially distributed on the stator connection surface (49); - determining a mesh of the geometry of the connecting surfaces (48, 49) and possibly of the respective blades (24, 25); the determination of the experimental design comprising intervals of authorized values for the parameters, sampling laws in these intervals, grid criteria and possibly a criterion for interrupting the simulation; the method then comprising a calculation phase (200) with the following steps: - exporting (100 ′, 204 ’) of the mesh to fluid mechanics calculation software, CFD; - simulating (201) the flow of air into the compressor stage (34, 134) and calculating the efficiency of the compressor stage; and - as long as a possible interruption criterion is not reached (202), the variation (203) of the values of one or more parameters in accordance with the sampling law, the definition of a mesh (204 ) corresponding and the iteration of the calculation; the method then comprising a phase of choosing (300) one or more geometries of connecting surfaces according to the results obtained during the calculation phase (200). 2. Method according to claim 1, characterized in that the parameters defining the splines (82, 84, 86) of the rotor support (26) comprise the coordinates of two respective points (70, 72, 74, 76, 78, 80) through which pass a given spline (82, 84, 86) and the parameters defining the splines (83, 85, 87) of the stator support (32) include the coordinates of two respective points (71, 73, 75, 77, 79, 81) through which a given spline passes (83, 85, 87). 3. Method according to one of claims 1 or 2, characterized in that in a given inter-blade space, the splines (82-87) are circumferentially spaced and are three in number (82, 84, 86) for the rotor support (26) and three (83, 85, 87) for the stator shell (25). 4. Method according to one of claims 1 to 3, characterized in that the values of the authorized parameters are such that the splines (82-87) do not move away radially from an axisymmetric portion (50, 51) of the bonding surface (48, 49) of more than 8% of the chord length of the respective blades. 5. Method according to one of claims 1 to 4, characterized in that the mesh criteria include limits imposed on the characteristics of the mesh, such as orthogonality, mesh angles, expansion rate. 6. Method according to one of claims 1 to 4, characterized in that the mailing criteria comprise admissible values of characteristics of the mesh, such as orthogonality, mesh angles, expansion ratio, and the method. comprises a step of monitoring (205) the characteristics of the mesh and when the characteristics of the mesh deviate from admissible values, the simulation is not carried out for the sample of given parameters or the results obtained by the simulation are annotated with corresponding information. 7. Method according to one of claims 1 to 6, characterized in that after the calculation phase (200), a response surface is constructed (301) by interpolation of the discrete results from each of the iterations, representing the calculated efficiency. during the calculation phase as a function of the parameters and the step of choosing one or more geometries of connection surfaces comprises the determination and selection of one or more sets of parameters for which the response surface has a local extremum or global. 8. Method according to one of claims 1 to 7, characterized in that the succession of initialization (100), calculation (200) and choice (300) phases defines a first calculation session, the method continuing. by a second calculation session comprising a second initialization phase (100), a second calculation phase (200) and a second choice phase (300), the experimental design of the second calculation session being different from the plan experience of the first calculation session. 9. Method according to one of claims 1 to 8, characterized in that the interruption criterion is a predefined number of iterations or a predefined calculation time. 10. Method according to one of claims 1 to 9, characterized in that the simulation step further comprises calculating the total pressure and / or the flow rate at the roots of the blades and the samples giving a result below a threshold. preset total pressure and / or flow are rejected. 11. Rotor (12) of a compressor (4) for a turbomachine, comprising a rotor support (26) and an annular row of rotor blades (24) extending radially outwardly to the rotor support (26), the rotor support (26) defining a rotor bonding surface (48) between two circumferentially adjacent rotor blades (24), the rotor bonding surface (48) conforming to a rotor support (26) bonding surface geometry (48) directly obtained by the method dimensioning of one of claims 1 to 10. 12. Compressor rectifier (4) for a turbomachine, comprising a stator support (32) and an annular row of stator vanes (25) extending radially outwardly from the stator support (32), the stator support (32) defining a surface. stator connection (49) between two circumferentially adjacent stator vanes (25), the stator connection surface (48) conforming to a connection surface geometry (49) of the stator support (32) directly obtained by the sizing process of one of claims 1 to 10. 13. Compressor rectifier (4) for a turbomachine, comprising a stator support (32) and an annular row of stator vanes (25) extending radially outwardly from the stator support (32), the stator support (32) defining a surface. stator connection (49) between two circumferentially adjacent stator vanes (25), the connection surface (49) being provided with axisymmetric portions (51) and non-axisymmetric portions (53), the latter (53) comprising a bump ( 56) and a hollow (57) between two circumferentially adjacent vanes (25A, 25B), characterized in that the hump (56) of the stator support (32) is attached to the lower surface (45) of a stator vane (25A ) and extends axially over at least 75%, preferably over 100%, of the chord (CS) of the stator vanes (25) and the hollow (57) of the stator support (32) is disposed circumferentially at approximately equidistant from two vanes (25A, 25B) circumferentially adjacent. 14. Rectifier according to claim 12 or 13, characterized in that the top (62) of this bump is attached to or close to the lower surface (45) of the stator vane (25A). 15. Rectifier according to one of claims 12 to 14, characterized in that the hollow (57) is disposed axially at a distance from the leading edges (41) of the stator vanes (25) of approximately 75% of the chord. (CS) of the stator vanes (25). 16. Stage (34, 134) of a compressor (4) for a turbomachine, comprising a rotor (12) according to claim 11 and a stator according to one of claims 12 to 15, the stator vanes (26) of the stator being arranged axially. directly downstream of the rotor blades (32) of the rotor (12).