• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    The invention relates to a method for determining tolerance intervals (ITi) for a set of dimensions of a part, said tolerance intervals being defined so that a part whose dimensions are included in said intervals respects a set of constraints with a determined probability of failure, the method being characterized in that it comprises the following steps: generating (100) a separation function for each constraint to be respected, each function being able to indicate whether a given dimensioning of a part satisfied or not with the corresponding constraint, and build (200) iteratively a set of intervals of tolerance of bigger and bigger by testing, with each iteration, with the aid of the functions of separation, if the dimensions included in said intervals respect constraints with a probability of failure lower than the probability of failure.
    公开号:FR3033062A1
    申请号:FR1551490
    申请日:2015-02-20
    公开日:2016-08-26
    发明作者:Abdelkader Otsmane；Virgile Marguin；Nicolas Gayton
    申请人:SNECMA SAS；
    IPC主号:
    专利说明:

    [0001] FIELD OF THE INVENTION The invention relates to a method for optimally determining tolerance intervals for a set of dimensions of a part. The invention applies in particular to the qualification of complex geometry parts, particularly in the field of aeronautics. The invention applies in particular to the qualification of tolerance intervals for turbomachine blades. STATE OF THE ART When designing a mechanical part, it is conventional to determine, for all the dimensions of the part that one seeks to design, nominal values. These values are defined as guaranteeing optimal performance of the part vis-à-vis a set of design criteria of multiple nature, such as for example cost, mechanical strength, aerodynamic performance, mass, etc. These nominal values are therefore the optimum values prescribed for the dimensions of the parts to be manufactured. However, the manufacturing process necessarily generates geometric deviations with respect to these nominal values, according to dispersions that are difficult to control. It is then necessary to be able to know if the parts exhibiting such deviations are also admissible with respect to these criteria, and thus to define optimum ranges of variation with respect to the nominal values which make it possible to ensure with a given probability that by reducing manufacturing costs. These ranges of variation from nominal values are tolerance ranges. These intervals are defined such that a representative sample of produced parts whose dimensions are included in the set of tolerance intervals corresponding to all the dimensions of the part has a probability of failure - that is, say a probability of not satisfying at least one of the constraints - determined. The determination of the tolerance intervals is a constrained parametric optimization problem which amounts to minimizing a cost function which decreases when the tolerance zone size increases overall, in order to facilitate the production of the part and to reduce the size thereof. while ensuring that the sizing included in this tolerance zone has a probability of stress failure less than the maximum probability set by the designer. However, the evaluation that a given dimensioning of a part satisfies all the design constraints requires a mechanical modeling of the part and a large number of calculations to verify its behavior, for example mechanical, aerodynamic, etc. This is particularly true for parts of complex geometry and having to satisfy a large number of constraints, as is the case in particular with turbomachine blades. As a result, solving the parametric optimization problem to determine the optimal tolerance intervals is a very time-consuming process (and much more than just looking for nominal dimensions) because of the multiple iterations required and a cost calculation (due to simulation) high. Indeed, since the optimization needs to explore all the possibilities in terms of room dimensions, a large number of calculations will be carried out.
    [0002] There is therefore a need for a method of determining tolerance ranges for dimensions of a part that is less expensive in terms of computation time and compatible design times. PRESENTATION OF THE INVENTION The purpose of the invention is to overcome at least one of the disadvantages mentioned above. In particular, the object of the invention is to propose a method of determining optimum tolerance intervals for a set of dimensions of a part which reduces both the manufacturing costs and the calculation times used for the search of these intervals.
    [0003] In this regard, the subject of the invention is a method for determining tolerance intervals for a set of dimensions of a part, said tolerance intervals being defined so that a part whose dimensions are included in said intervals respects a set of constraints with a determined probability of failure, the method being characterized in that it comprises the following steps: generating a separation function for each constraint to be respected, each function being able to indicate whether a given dimensioning of one piece satisfied or not to the corresponding constraint, and - iteratively build a set of tolerance intervals of bigger and bigger by testing, with each iteration, using the functions of separation, if the sizing included in said intervals respect the constraints with a probability of failure lower than the probability p. Advantageously, but optionally, the method according to the invention further comprises at least one of the following features: the step of constructing the set of tolerance intervals comprises the iterative implementation of the steps of: a) globally increase the current tolerance intervals with respect to the previous tolerance intervals so that there are more sizings included in the current intervals than in the previous intervals, b) randomly draw according to a multi-variate Gaussian law centered on nominal dimensions of the dimensions of the part, a plurality of sizing included in the set of common tolerance intervals, c) testing, by means of the separating functions, whether each dimensioning is admissible with respect to the constraints, d) testing whether the percentage of dimensionings which are not admissible with respect to the constraints is lower than the probability p of failure launches, and e) in the case of a positive test in step d), repeat steps a) to d) 30 - during the first iteration of the step of construction of the tolerance intervals, each interval is reduced to a value determined nominal of the corresponding dimension of the piece. The method comprises, in the case of a negative test in step d), a step e ') of repeating steps a) to d), taking again as tolerance intervals 3033062 4 the last tolerance intervals for which the percentage of dimensionings that are not admissible with respect to the constraints is less than the probability p of failure. step a) is carried out by disturbing the ends of the preceding tolerance intervals symmetrically according to a Gaussian law. step a) further comprises, after the disturbance, the calculation of the total volume of a hypercube whose sides are constituted by the sizes of the tolerance intervals, and the comparison of the volume of the hypercube relative to to that constituted by the last intervals for which the percentage of dimensionings which are not admissible with respect to the stresses is less than the probability p of failure. if the volume of the hypercube obtained is less than that constituted by the last intervals for which the percentage of dimensionings which are not admissible with respect to the stresses is less than the probability p of failure, step a) is repeated by taking up again as previous tolerance intervals, the last tolerance intervals for which the percentage of dimensionings which are ineligible for the constraints is less than the probability p of failure. the step of constructing the tolerance intervals stops after a determined number of unsuccessful iterations defined by: the volume of the current hypercube (Vk) is smaller than the volume (Vko) of the last hypercube corresponding to the intervals of tolerances for which the number of non-allowable sizing under the constraints is less than the probability p of failure, or o the result of step d) is negative. each constraint is modeled by a function, and the step of generating (100) separating functions comprises the implementation of the following steps: constituting a scaling Monte-Carlo population of the part (110), o initializing, for each constraint (g,), a corresponding Kriging separator (120), and iteratively implementing the steps of: o determining, (130) among the sizing of the population of Monte-Carlo, the one for which the precision of estimation of a constraint by the corresponding Kriging separator is minimal, and 5 o compute, (140) for this dimensioning, the value of the set of constrained functions, and set day the Kriging separators accordingly. - the piece is a turbomachine dawn.
    [0004] The invention also relates to a computer program product, comprising code instructions for carrying out the method according to the foregoing description, when executed by a processor. The invention finally relates to a method of manufacturing a turbomachine part, comprising, for each manufactured part: the measurement of all the dimensions of the part, the verification that each dimension is included in an interval corresponding tolerance, in which the tolerance intervals are determined by the implementation of the method according to the foregoing description, and - if each dimension of the part is included in the corresponding tolerance interval, the validation of the part, and if at least one dimension is not included in the corresponding tolerance interval, performing complementary processing on the part to determine whether the part is valid or not. The proposed method solves the aforementioned problem of reducing the calculation time during the development of tolerance intervals by minimizing the number of modelizations of a part to check its compliance with the constraints.
    [0005] Indeed, these modelizations, which are the heaviest steps in computing time, are implemented only to define the separation functions for each constraint to be respected. Once these separation functions are generated, it is possible to test by a faster calculation if a given dimensioning satisfies the constraints, and this with precision.
    [0006] 3033062 6 These separation functions therefore make it possible to test a large number of dimensions in a limited time, and thus maximize the tolerance intervals for a part. The proposed method advantageously applies to parts of complex geometry comprising a mean number, or even a large number of geometric dimensions and constraints to be respected, such as for example turbomachine blades. DESCRIPTION OF THE FIGURES Other features, objects and advantages of the invention will emerge from the description which follows, which is purely illustrative and nonlimiting, and which should be read with reference to the appended drawings, in which: FIG. sectional view of the profile of a turbomachine blade, with some characteristic dimensions that can be used in the parametric optimization of the tolerance intervals. FIG. 1b is a table giving an example of design dimensions of a turbomachine blade, at different heights of the blade (0%, 20%, 30%, 60%, 70%, 85%, 100%). FIG. 2 represents the main steps of a method for determining tolerance intervals according to one embodiment of the invention. FIGS. 3a to 3d show schematically the implementation of the step of construction of the tolerance intervals of the method of FIG. 2. DETAILED DESCRIPTION OF AT LEAST ONE MODE OF IMPLEMENTING THE INVENTION Referring to FIG. FIG. 1a shows an example of a part and corresponding design variables, for which tolerance ranges are sought. Design variables can be measures, for example length, width, thickness, height, and / or angles. Subsequently these variables are qualified under the generic term of 30 dimensions, a set of values given for each design variable constituting an example of sizing of the part. In the non-limiting example of Figure la the piece is a turbomachine blade, sectional view at a given height.
    [0007] FIG. 1b gives an example of dimensions for which it is desired to determine tolerance intervals. These dimensions are for example a maximum thickness, a rope dimension (that is to say the segment connecting the leading edge to the trailing edge of a blade) and a wedging angle (the angle between the rope 5 and the axis of revolution of the turbine engine in which is positioned the blade) of the blade, for sections at different heights of the blade, the number associated with each variable in Figure 1 b designating the height of the section of the dawn as a percentage of the total height. According to this example, there are therefore 13 dimensions each associated with a determined tolerance interval. In addition, each dimension is associated with a nominal value obtained at the end of a design process that does not take into account the variability of the dimensions inherent to the manufacturing process. In the following, the dimensions for which the 15 optimum tolerance intervals are sought, i being an index varying between 1 and N, where N is the number of design dimensions of a part (N = 13 according to the example). previous), xliwm the corresponding nominal values, and 17'i = rxmmW, xlinax] the corresponding tolerance intervals.
    [0008] Method for Determining Tolerance Intervals A method for determining the tolerance intervals 17'i for a set of x-dimensions of a part will now be described. This method is implemented by a processing unit (not shown), for example a computer, comprising processing means, for example a processor, configured to execute an appropriate program. The tolerance intervals are defined 17'i so that the sizing of parts included in all the intervals must respect a set of n constraints with a probability of failure p fixed by the designer, the probability of failure being defined as the probability that sizing does not satisfy at least one of the constraints. Taking again the preceding example of a turbomachine blade, the set of constraints to be respected can for example be defined as follows: - Mass less than a determined maximum mass, 3033062 8 - Displacement of a point induced by forces in operation less than a determined threshold displacement, - Rigidity greater than a certain value, - Aerodynamic constraints, 5 - Set of mechanical stresses lower than a set of threshold values (for example a maximum Von Mises stress at the top of the blade), - Frequencies dynamic in an acceptable range, non-coinciding with motor harmonics, etc.
    [0009] The probability p (therefore between 0 and 1) is set by the designer, preferably less than 0.1, and preferably less than 0.01. Each constraint is associated with a mathematical function modeling it, named g 'gi: x E RN R (i = 1, ..., n), characterized by a limit state gi = 0 corresponding to a border beyond which a sizing no longer satisfies the corresponding constraint. For example, a function g, evaluated in a first dimensioning is negative if the dimensioning does not respect the constraint, and positive if the dimensioning respects it. The objective is therefore to optimize the tolerance intervals with a view to minimizing production costs. This can be reformulated mathematically as follows: Find (1T1, ..., IT,), such that: (17'1, ..., / TO = argmin (f (/ T1, ..., / 7 ,, )) {p <p Where: - f (17'1, ..., 17.2) denotes a cost function associated with tolerance intervals, i.e. a function to be minimized (usually this function is inversely proportional to the tolerance intervals), for information, it can be expressed in the form: f (Irri, ..., Un) = ai * ITi i = t and ai is a weighting coefficient, for example the sum of ai may be equal to 1, 3033062 9 - / T denotes a measure of the tolerance interval (normalized) associated with each dimension, - p is the probability of not complying with at least one of the constraints posed (under hypothesis the constraints are independent) 5 - po is a maximum failure probability not to be exceeded The method for determining the tolerance intervals, summarized schematically in Figure 2, comp The first step is a step 100 of generating a set of 10 separating functions, comprising as many separation functions as there are constraints to be respected. Each separation function makes it possible to indicate, for a given dimensioning, whether this dimensioning satisfies the constraint associated with the separation function, that is to say that it makes it possible to classify the points of RN on one side or the other of the limit state defined by gi = 0, for each constrained function g 'and this without calculation cost contrary to the conventional evaluation, by mechanical calculations, that a part satisfies these constraints. According to a preferred embodiment of step 100, the separation functions are Kriging separators, generated by implementing a so-called AK-MCS strategy described in the publication by B. Echard, N. Gayton and M. Lemaire. : "AK-MCS: An Active Learning Reliable Methodology Combining Kriging and Monte Carlo Simulation", Structural Safety, vol. 33, pages 145-154, 2011, and which includes the following substeps. Step 100 first includes a population initialization 110 for generating Kriging separators. During this initialization 110, study windows [Xmin, Xm "] are defined for each dimension x,. A study window corresponds to the largest interval in which a dimension x 'can be understood independently of the respect of the constraints g ,.
    [0010] Step 110 then comprises generating a Monte Carlo population according to a uniform scaling law (x1, ..., xN) of which each element is included in the corresponding study window (the size of this population being of the order of tens of thousands of points).
    [0011] Step 100 then comprises a step of initializing Kriging separators 120. This step firstly comprises the evaluation of each constraint function g, for a set of included dimensions, each dimension of which is included in the window of corresponding study.
    [0012] In this respect, the set of sizing may come from an initial design plan of the designer of the object to be dimensioned (for example a Latin Hypercube type plane of given size depending on the number of dimensions involved). that is to say a set of k dimensioningsj = 1..k already available to the designer.
    [0013] In a variant, the set of sizings is generated from the Monte Carlo population generated in step 110. A set of k vectors (4 :, such that each xi) is then randomly drawn within this population. is included in the corresponding study window Peinin, Xinax], k being of the order of 5 to 10 times the number of dimensions.
    [0014] Then, the constraints g, are calculated in these k vectors, by direct call to the simulation models. This then makes it possible to perform a kriging of the values of the constrained functions g, on the set of sizing of the Monte-Carlo population from the points xi for which the constraints have been calculated. Kriging is a known interpolation method, described for example in G. Matheron's publication, "The intrinsic random functions and their applications", Adv Appl Probab 1973; 5 (3): 439-68. At the end of this step 120 a kriging separator Ili is obtained for each constraint, the separator indicates an estimate of the value of the corresponding stress g, in a given dimensioning. However, given the small number of points used in the development of separators, they are of rather poor quality and do not allow to determine accurately whether a sizing meets the constraints or not. The following steps of the process are aimed at improving the quality of the separators. To do this, steps 130 and 140 are implemented iteratively.
    [0015] Step 130 comprises, for each sizing x1 of the Monte Carlo population, the evaluation of the set of separators relating to the constraints g, to deduce therefrom, for each of them: the predictor, or Kriging separator, i.e., the estimate of the value of g, e), and the variance o-i2 (xi) of Kriging, i.e. the minimum of the error quadratric mean between Ili (xi) and gi (x1). These elements are calculated in accordance with the publication cited above, in which the predictor is Ô (x).
    [0016] Once these elements are calculated, we deduce the criterion Ui () =) 1/0-i2 () for each constraint g ,. For each individual x1 E RN of the Monte-Carlo population, we retain only the criterion Ui (xl) of the constraint providing the prediction Ili (xi) the lowest in absolute value (the constraints having previously been normalized): note U (xi) 15 this criterion. Then, among all the points of the population of Monte Carlo, we retain only the point xi providing the weakest criterion U, signifying the greater probability of being mistaken for a sign at this point. In step 140, all constraints are calculated at point x1 determined at the end of step 130, and the Kriging separators are updated accordingly. At each iteration the Kriging separators are thus enriched with additional information and made more precise. This enrichment, implemented in steps 130 and 140, is iterated until 99% of the U values for the remaining population of the uniform Monte Carlo pull are greater than 2. These empirical values are derived from experiments carried out during the validation of the AK-CMS method of the publication cited above and allow an optimal classification of the points of the x1 population vis-à-vis the admissibility boundary of each constrained function g, = 0. At the end of step 100, this results in a representative population among all the sizing possible in the study windows for which the Kriging predictor is evaluated and of which we therefore have a very good estimate of the respect constraints. Step 100 is relatively expensive in computation time because of the implementation, in steps 120 and 140, of computations using the simulation models to accurately calculate the values of the constraints in a sizing. However, this step then makes it possible to define the largest tolerance intervals without additional expensive calculations, which in turn make it easy to determine the eligible parts and parts requiring additional verification, as indicated below. The method for determining the tolerance intervals then comprises a second main step 200 using the separation functions determined in step 100.
    [0017] This step 200 is a step of iteratively building tolerance intervals minimizing a cost function associated with part manufacturing costs (this cost being reduced with larger tolerance intervals), by testing, at each iteration, if a number of sizings randomly drawn in these tolerance ranges satisfy the constraints with a failure probability lower than the determined probability p. The set of tolerance intervals 1T1, i = 1, .., N forms a hypercube of dimension N defined by its center, in this case the dimension vector N of nominal design dimensions (xrm, .., xien) and by a vector of dimension N (ti, .., tN) defining the size of the hypercube following each variable x ,.
    [0018] The implementation of step 200 is represented in a simplified manner in FIGS. 3a to 3d in dimension 2, that is to say in the case of a two-dimensional part, and by choosing two constraints to be respected. . The curves shown schematically illustrate the separators to be respected, and the clear frame in the center represents the study area formed by the study windows for the two dimensions (i.e. the tolerance intervals). With reference to FIG. 3a, step 200 is initialized with, for each dimension, tolerance intervals reduced to the nominal values xliwm. Then, step 200 includes the iterative implementation of the following steps.
    [0019] With reference to FIG. 3b, during a substep 210, each preceding tolerance interval - that is to say, resulting from a previous iteration or, in the case of a first iteration, the corresponding nominal value - is disturbed randomly and symmetrically - in order to maintain the nominal dimensions as the central point of the hypercube, as shown in FIG. 3b - according to a Gaussian law. The law is centered with a small standard deviation before the size of the admissible domain by the constraint functions. According to an advantageous embodiment, the disturbances along the directions of the design space are normalized so as not to favor a particular dimension, and the perturbation law is of normal type with a very low perturbation, for example of 0 , 1, that is to say that the distance between a corner of the initial hypercube and the same corner of the disturbed hypercube is of the order of 0.1 in the space dimensioned variables. During the first iteration, the tolerance intervals being reduced to 15 nominal values, it is these values which are disturbed according to the Gaussian law, until forming for each dimension an interval of length t determined. During the following iterations, it is the ends xlinin and xlinax which are disturbed, the perturbation corresponding to a normed growth of the hypercube of small amplitude in the case of the embodiment presented above.
    [0020] At the end of this step, a hypercube of sides t is obtained, representing the length of the tolerance intervals obtained. This hypercube is represented in the simplified example of FIG. 3b as a side rectangle t1 for the interval corresponding to the first dimension and t2 for the interval corresponding to the second dimension.
    [0021] The hypercube obtained must be larger and larger with the iterations, since it is sought to increase the tolerance intervals so as to minimize the cost function of the optimization problem. In other words, at each iteration, there must be more sizing included in the current tolerance intervals than in the previous intervals for which the probability of failure is respected. To verify this, step 200 includes a check 220 that the hypercube obtained at the end of step 210 is larger than the last previously built admissible hypercube. This step 220 takes place only from the 2nd iteration of the algorithm.
    [0022] By admissible hypercube is meant a hypercube corresponding to noted tolerance intervals / Tko for which the percentage of dimensionings which are not admissible with respect to the stresses, calculated during a step 250 described hereinafter of a preceding iteration, is less than the probability p 5 of failure. It can be the hypercube obtained at the previous iteration but not necessarily, if this hypercube included a too large percentage of inaccessible dimensionings. To carry out this verification 220, the volume Vk of the current hypercube 10 is calculated and compared with the volume Vko of the last acceptable hypercube with: Vk = ti, k where t ,, k is the length of the tolerance interval 17 ' i current, that is to say of the iteration k. In the case where the current hypercube has a smaller volume than the hypercube to which it has been compared, step 210 is reiterated from the last admissible hypercube. In the case where the current hypercube has a greater volume than the preceding hypercube, the method then comprises step 230 during which points are drawn randomly in the hypercube, according to a multivariate Gaussian law centered on the values nominal dimensions. The 20 points drawn randomly are as many sizing of the piece. Advantageously, the standard deviations of the dimensions x, forming the elements of the variance-covariance matrix of the Gaussian law used for the draw, are respectively equal to t / 6, in accordance with the commonly accepted 6-sigma approach.
    [0023] The number of sizing drawn is at least equal to 10n + 2 to estimate a probability of failure of the order of 10-n. Once the sizing is drawn randomly, the method comprises a step 240 of testing, for each dimensioning, whether it meets all the constraints to be met.
    [0024] To do this, the evaluations of each dimensioning are recovered with each of the separation functions respectively corresponding to each of the constraints to be respected, that is to say Kriging separators, stored in step 100. This step is therefore of rapid implementation since it does not require to generate a model of the part according to each of the 5 dimensions, before implementing mechanical calculations to verify the respect of constraints. A dimensioning is considered admissible if it satisfies all the constraints, and inadmissible if it does not satisfy at least one constraint. Once all the sizing has been tested, the method comprises a step 250 counting allowable sizing among the set of tested sizing, and comparing the percentage of inadmissible sizing with the probability p of failure to meet. If the percentage of inadmissible sizing is less than the required failure probability p, then the hypercube constituted by the current tolerance intervals is considered permissible. In this case, the steps 210 to 250 are reproduced, taking as previous tolerance intervals those obtained at the end of this step 250, as shown schematically in FIGS. 3c and 3d (enlargement of the preceding hypercube to obtain the current hypercube).
    [0025] On the other hand, if at the end of step 250 the percentage of inadmissible sizing is greater than the probability p of failure, then steps 210 to 250 are reproduced from the last hypercube for which the percentage of inadmissible sizings is less than to the probability p of failure - corresponding to the intervals / Tko - or, in the case where there was no other iteration, by taking the initial nominal values. Steps 210 to 250 are reproduced iteratively until a determined number of unsuccessful iterations are obtained, an unsuccessful iteration being defined by: either the current hypercube is smaller than the preceding hypercube, or inadmissible sizing is greater than the probability p. Advantageously, the method comprises at least 50, for example 100 unsuccessful iterations before stopping, this number of iterations increasing the probability of having obtained the optimum tolerance intervals.
    [0026] The proposed method thus allows each step to increase the size of the tolerance intervals while quickly verifying that sizing selected in these intervals have a lower probability than not satisfying one of the constraints. This results in maximized intervals ensuring compliance with the probability of failure imposed. Once the tolerance intervals obtained for all the dimensions of the part, the manufacture of the part uses these tolerance intervals by measuring, for each piece manufactured, all of its dimensions, and then checking whether each of the dimensions is included in the corresponding tolerance range. If each dimension of a part is included in the corresponding tolerance range, then the part is considered valid. If one or more dimensions of the part are not included in the corresponding tolerance range, this does not necessarily imply that the part is invalid. To determine it, the part is redirected to a specific treatment during which complementary calculations are carried out. The tolerance ranges defined above make it possible to achieve significant savings in production, since a larger number of pieces is determined to be statistically valid during manufacture, and the processing involving the complementary calculation is implemented to a smaller number of rooms.
    权利要求:
    Claims (12)
    [0001]
    REVENDICATIONS1. A method for determining tolerance intervals (IT,) for a set of dimensions (x,) of a part, said tolerance intervals (IT,) being defined so that a part whose dimensions (x,) are included in said intervals respects a set of constraints with a given probability p of failure, the method being implemented by a processing unit and being characterized in that it comprises the following steps: generating (100) a separation function (p,) for each constraint (g,) to be respected, each function being able to indicate whether a given dimensioning (x1, ..., xN) of a part satisfies the corresponding constraint or not, and - to construct (200) ) iteratively a set of tolerance intervals (IT,) becoming larger by testing, at each iteration, using the separation functions, if the sizing (x1, ..., xN) included in said intervals respect the constraints with a pr failure reliability less than probability p.
    [0002]
    The method of claim 1, wherein the step (200) of constructing the set of tolerance intervals (11-i) comprises iteratively implementing the steps of: a) magnifying (210) globally the current tolerance intervals (ITi, k) with respect to the previous tolerance intervals so that there are more sizings included in the current intervals (ITi, k) than in the previous intervals (IT ,, ki), b) randomly drawing (230) according to a multi-variate Gaussian law centered on nominal values of the dimensions of the part, a plurality of dimensions included in the set of common tolerance intervals (ITI, k), c) testing (240), using the separation functions, if each dimensioning is admissible with respect to the constraints, d) test (250) whether the percentage of dimensionings that are not admissible with respect to the stresses is less than the probability p of failure, and 3033062 18 e) in positive test case in step d), repeat steps a) to d)
    [0003]
    3. A method according to claim 2, wherein in the first iteration of the tolerance step (200), each interval is reduced to a given nominal value (x ') of the corresponding dimension of the coin. .
    [0004]
    4. Method according to one of claims 2 or 3, comprising, in the case of a negative test in step d), a step e ') of reiterating steps a) to d) by taking again as intervals of previous tolerances (1-ri, k_i) the last tolerance intervals (ITi, ko) for which the percentage of dimensionings that are not admissible with respect to the constraints is less than the probability p of failure.
    [0005]
    5. Method according to one of claims 2 to 4, wherein step a) is implemented by disturbing the ends of tolerance intervals (IT ,, k_i) previous symmetrically according to a Gaussian law.
    [0006]
    The method of claim 5, wherein step a) further comprises, after the disturbance, calculating (220) the total volume (Vk) of a hypercube whose sides are constituted by the sizes. tolerance intervals, and the comparison of the volume (Vk) of the hypercube relative to that (Vko) constituted by the last intervals (IT ,, k0) for which the percentage of dimensionings that are not admissible with respect to the stresses is less than the probability p of failure. 25
    [0007]
    7. The method of claim 6, wherein if the volume of the obtained hypercube is less than that constituted by the last intervals (ITi, k0) for which the percentage of dimensionings ineligible for the constraints is less than the probability p step a) is reiterated by repeating as previous tolerance intervals (ITi, k_i) the last tolerance intervals (IT ,, ko) for which the percentage of dimensionings that are not permissible under the constraints is less than the probability p of failure. 3033062 19
    [0008]
    The method according to claim 7, wherein the step of constructing the tolerance intervals stops after a determined number of unsuccessful iterations defined by: the volume of the current hypercube (Vk) is less than the volume (Vko) ) of the last hypercube corresponding to the tolerance intervals for which the number of non-allowable sizing under constraints is less than the probability p of failure, or the result of step d) is negative. 10
    [0009]
    9. Method according to one of the preceding claims, wherein each constraint is modeled by a function, and the step of generating (100) separation functions comprises the implementation of the steps of: - constitute a population of Monte -Carlo sizing of the part (110), 15 - initialize, for each constraint (gi), a corresponding Kriging separator (120), and the implementation, iteratively, steps of: - determine, (-) 130) among the sizing of the Monte Carlo population, that for which the accuracy of estimation of a constraint by the corresponding Kriging separator is minimal, and - calculating, (140) for this dimensioning, the value of the set of constraint functions, and update the Kriging separators accordingly. 25
    [0010]
    10. Method according to one of the preceding claims, wherein the part is a turbomachine blade.
    [0011]
    A computer program product, comprising code instructions for carrying out the method according to one of the preceding claims, when executed by a processor.
    [0012]
    12. A method of manufacturing a turbomachine part, comprising, for each manufactured part: the measurement of all the dimensions of the part, the verification that each dimension is included in a corresponding tolerance range, in which tolerance ranges are determined by the implementation of the method according to one of claims 1 to 10, and 5 - if each dimension of the part is included in the corresponding tolerance interval, the part validation, and if at least one dimension is not included in the corresponding tolerance interval, the implementation of a complementary processing on the part to determine whether the part is valid or not. 10
    类似技术:
    公开号 | 公开日 | 专利标题
    EP2987102B1|2022-01-26|Method of modeling a workpiece, in partucilar a turbine blade
    FR3033062A1|2016-08-26|METHOD FOR DETERMINING TOLERANCE INTERVALS FOR SIZING A WORKPIECE
    EP3472736A1|2019-04-24|Method for estimating stress intensity factors and method for calculating associated service life
    WO2007101939A1|2007-09-13|Method for characterizing the endurance limit of a part from its surface profile
    WO2014064396A2|2014-05-01|System for monitoring a set of components of a device
    Park et al.2018|Data compression and prediction using machine learning for industrial IoT
    WO2015036390A1|2015-03-19|Computer-assisted design method comprising a modelling step
    FR2944125A1|2010-10-08|METHOD FOR IDENTIFYING AERODYNAMIC MODELS FOR AIRCRAFT SIMULATION METHOD
    EP3028193B1|2019-04-17|Method for modelling an open rotor blade
    EP2846280A1|2015-03-11|Mechanical strain gauge simulation
    EP2780665B1|2020-02-12|Design for a part made from 3d woven composite material
    WO2014001070A1|2014-01-03|Device and method for optimising a fuzzy inference system maintaining the interpretability
    EP2904587A2|2015-08-12|Method for processing data defining an element in a space e of dimensions d, and associated computer program
    EP3757906A1|2020-12-30|Automatic determination of parameters for executing a software application on a platform for processing information by genetic algorithm and improving noise management
    FR3064387A1|2018-09-28|METHOD FOR MANUFACTURING PARTS BASED ON STABILITY ANALYSIS OF A CHARACTERISTIC DIMENSION
    FR3027133A1|2016-04-15|METHOD FOR MODELING A BLADE OF A NON-CARBENE PROPELLER
    JP5211656B2|2013-06-12|Toroidal surface evaluation method
    US11263172B1|2022-03-01|Modifying a particular physical system according to future operational states
    EP3622445A1|2020-03-18|Method, implemented by computer, for searching for rules of association in a database
    EP2405373A2|2012-01-11|Method and system for modelling a geometric surface
    FR3029316A1|2016-06-03|METHOD FOR MAKING A CARTOGRAPHY OF SIGNALS
    CN112215398A|2021-01-12|Power consumer load prediction model establishing method, device, equipment and storage medium
    FR3105863A1|2021-07-02|Method AND system for designing a prediction model
    FR3023634A1|2016-01-15|COMPUTERIZED SYSTEM FOR DESIGNING THE THREE-DIMENSIONAL ROUTING OF ELECTRIC CABLES IN AN ELECTRICAL SYSTEM, AND CORRESPONDING DESIGN PROCESS
    FR3075995A1|2019-06-28|METHOD FOR DETECTING ABERRANT DATA.
    同族专利:
    公开号 | 公开日
    US20160246917A1|2016-08-25|
    FR3033062B1|2021-09-03|
    US10296705B2|2019-05-21|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    FR2669420B1|1990-11-21|1993-01-15|Hispano Suiza Sa|METHOD FOR CONTROLLING DIMENSIONAL MEASUREMENTS OF FOUNDRY PARTS.|
    JP5024017B2|2007-12-14|2012-09-12|富士通株式会社|Tolerance analysis calculation system, tolerance analysis method and program|
    US8447549B2|2011-02-11|2013-05-21|Quality Vision International, Inc.|Tolerance evaluation with reduced measured points|
    US9558300B2|2011-11-11|2017-01-31|Carnegie Mellon University|Stochastic computational model parameter synthesis system|CN107679315B|2017-09-28|2021-08-10|上海交通大学|Geometric compensation method and system for welding deformation of vehicle body plate|
    US11250189B1|2018-04-30|2022-02-15|United States Of America As Represented By The Secretary Of The Air Force|Identifying effects of geometric variations of physical parts|
    CN112163331A|2020-09-24|2021-01-01|广东电网有限责任公司电力科学研究院|Distribution network line vulnerability calculation method and related device|
    法律状态:
    2016-02-08| PLFP| Fee payment|Year of fee payment: 2 |
    2016-08-26| PLSC| Publication of the preliminary search report|Effective date: 20160826 |
    2017-02-01| PLFP| Fee payment|Year of fee payment: 3 |
    2018-01-23| PLFP| Fee payment|Year of fee payment: 4 |
    2018-02-02| CD| Change of name or company name|Owner name: SAFRAN AIRCRAFT ENGINES, FR Effective date: 20170719 |
    2020-01-22| PLFP| Fee payment|Year of fee payment: 6 |
    2021-01-20| PLFP| Fee payment|Year of fee payment: 7 |
    2022-01-19| PLFP| Fee payment|Year of fee payment: 8 |
    优先权:
    申请号 | 申请日 | 专利标题
    FR1551490A|FR3033062B1|2015-02-20|2015-02-20|PROCESS FOR DETERMINING TOLERANCE INTERVALS FOR THE SIZING OF A PART|FR1551490A| FR3033062B1|2015-02-20|2015-02-20|PROCESS FOR DETERMINING TOLERANCE INTERVALS FOR THE SIZING OF A PART|
    US15/048,331| US10296705B2|2015-02-20|2016-02-19|Method for determining tolerance intervals for dimensioning a part|
    [返回顶部]