A method for tolerance allocation of unit performance based on multi-level performance transfer

Abstract

The present application provides a method for tolerance allocation of unit performance based on multi-level performance transfer, belonging to the technical field of aero-engines, including constructing a forward proxy model of the whole machine containing multiple unit bodies based on geometric feature parameters and the whole machine performance parameters; in this model, for each unit body, identify the key geometric feature parameters, reduce the degrees of freedom of the feature distribution of each key geometric feature parameter to obtain a single parameter; input the single parameter and the numerical range corresponding to each key geometric feature parameter into the forward proxy model of the whole machine for updating; for the whole machine target performance requirements, perform uncertainty backward propagation based on the updated forward proxy model of the whole machine to obtain the geometric feature parameter distribution of each unit body, and based on the obtained distribution, use the trade-off space to obtain the tolerance allocation result. The solution of the present application can ensure the performance consistency and interchangeability of unit bodies, and promote the agile R & D, rapid derivation, and condition-based maintenance of aero-power.

Description

Technical Field

[0001] This application relates to the technical field of aeroengines, and particularly to a method for allocating unit performance tolerances based on multi-level performance transmission. Background Art

[0002] Traditional deterministic designs ignore inherent uncertainties, so a large degree of conservatism is required in engine design to reserve sufficient margins to prevent performance from falling short. In the prior art, no method has been proposed to avoid excessive conservatism and further reduce the margins to optimize the margin requirements in engine design. Summary of the Invention

[0003] In view of this, embodiments of this application provide a method for allocating unit performance tolerances based on multi-level performance transmission, which at least partially solves the problem of excessive conservatism in margins in engine design existing in the prior art.

[0004] Embodiments of this application provide a method for allocating unit performance tolerances based on multi-level performance transmission, and the method includes:

[0005] Construct a forward proxy model of the whole engine including multiple units based on geometric feature parameters and the whole engine performance parameters;

[0006] In the forward proxy model of the whole engine, for each unit, identify the key geometric feature parameters;

[0007] For each unit, perform dimensionality reduction of the degrees of freedom of the feature distribution of each key geometric feature parameter to obtain a single parameter;

[0008] For each unit, input the single parameter corresponding to each key geometric feature parameter and the numerical range corresponding to each key geometric feature parameter into the forward proxy model of the whole engine for updating;

[0009] For the whole engine target performance requirements, perform uncertainty backward propagation based on the updated forward proxy model of the whole engine to obtain the geometric feature parameter distribution of each unit;

[0010] According to the obtained geometric feature parameter distribution of each unit, use the trade-off space to obtain the tolerance allocation result.

[0011] According to a specific implementation manner of embodiments of this application, the constructing a forward proxy model of the whole engine including multiple units based on geometric feature parameters and the whole engine performance parameters includes:

[0012] For each unit, randomly select multiple samples from various geometric feature parameter degradation probabilities, perform CFD simulations on all different sample combinations, and calculate the component characteristics under the influence of geometric feature parameter degradation;

[0013] Input the component characteristics obtained from the simulation of each unit cell into a zero-dimensional overall machine performance model to obtain the overall machine performance under the combined action of multiple geometric feature parameter degradation modes;

[0014] Based on the obtained multiple overall machine performances, construct an overall machine forward proxy model based on geometric feature parameters and overall machine performance parameters through the proxy model method.

[0015] According to a specific implementation manner of an embodiment of the present application, in the overall machine forward proxy model, for each unit cell, identify key geometric feature parameters, including:

[0016] Perform sensitivity analysis based on the overall machine forward proxy model, and calculate the SHAP value of each geometric feature parameter in each unit cell respectively;

[0017] For each unit cell, sort the geometric feature parameters according to the SHAP value, and identify the key geometric feature parameters.

[0018] According to a specific implementation manner of an embodiment of the present application, perform dimensionality reduction of the degrees of freedom on the feature distribution of each key geometric feature parameter to obtain a single parameter, including:

[0019] Represent the first four statistical moments (μx; σx; Sx; Kx) of the probability density of each key geometric feature parameter by a single parameter to complete the dimensionality reduction of the degrees of freedom; where μx is the mean, σx is the standard deviation, Sx is the skewness, and Kx is the kurtosis.

[0020] According to a specific implementation manner of an embodiment of the present application, perform dimensionality reduction of the degrees of freedom on the feature distribution of each key geometric feature parameter to obtain a single parameter, including:

[0021] When the feature distribution of the geometric feature parameter is a triangular distribution, define the single parameter dx as:

[0022] dx = x ub -x lb ,

[0023] where x ub is the upper boundary of the triangular distribution, and x lb is the lower boundary of the triangular distribution.

[0024] According to a specific implementation manner of an embodiment of the present application, input the single parameter corresponding to each key geometric feature parameter and the numerical range corresponding to each key geometric feature parameter into the overall machine forward proxy model for update, including:

[0025] Construct a mixture Gaussian distribution based on the single parameter corresponding to each key geometric feature parameter and the numerical range corresponding to each key geometric feature parameter;

[0026] Input the Gaussian mixture distribution into the overall machine forward proxy model for updating.

[0027] According to a specific implementation manner of an embodiment of the present application, for the overall machine target performance requirements, perform uncertainty backpropagation based on the updated overall machine forward proxy model to obtain the geometric feature parameter distribution of each unit body, including:

[0028] For the overall machine target performance requirements, calculate the target performance standard deviation σ according to the confidence interval Ay ;

[0029] In each unit body, use Monte Carlo sampling to extract multiple geometric feature parameters from the empirical distribution to form the standard deviation of multiple geometric feature parameter distributions;

[0030] Input the standard deviations of multiple geometric feature parameter distributions into the updated overall machine forward proxy model, and calculate the standard deviation σ of the simulated distribution of the overall machine performance Ty ;

[0031] When the standard deviation σ of the simulated distribution of the overall machine performance Ty and the target performance standard deviation σ Ay satisfy the preset matching requirements, then accept the standard deviation σ of the simulated distribution of the overall machine performance Ty ;

[0032] When the difference between the standard deviation σ of the simulated distribution of the overall machine performance Ty and the target performance standard deviation σ Ay does not satisfy the preset matching requirements, then continue to adjust the multiple geometric feature parameter distributions for iterative optimization until the preset matching requirements are met;

[0033] Obtain the multiple geometric feature parameter distributions corresponding to when the preset matching requirements are met.

[0034] According to a specific implementation manner of an embodiment of the present application, the tolerance allocation result is obtained by using the trade-off space according to the obtained geometric feature parameter distribution of each unit body, including:

[0035] Convert the obtained geometric feature parameter distribution of each unit body into a trade-off space, and visually display the tolerance allocation result through the two-dimensional cross-sectional view of the trade-off space.

[0036] According to a specific implementation manner of an embodiment of the present application, the visual display of the tolerance allocation result through the two-dimensional cross-sectional view of the trade-off space includes:

[0037] Decompose the n-dimensional input space of the trade-off space into parameter pairs, where each parameter pair consists of two different geometric feature parameters. The n-dimensional input space is composed of the geometric feature parameters of all obtained unit bodies, and each parameter pair corresponds to a two-dimensional cross-section of a hypercube in the trade-off space;

[0038] Perform contour plotting on each two-dimensional cross-section to obtain the m-dimensional output space of the trade-off space, and visually display the tolerance allocation structure through the m-dimensional output space. Among them, the m-dimensional output space is composed of the overall machine performance parameters, and each contour line corresponds to a constraint on the deterministic or probabilistic geometric feature parameters of an overall machine performance parameter.

[0039] According to a specific implementation manner of the embodiment of the present application, the overall machine forward proxy model is constructed based on a machine learning method.

[0040] Beneficial effects:

[0041] In the unit body performance tolerance allocation method based on multi-level performance transfer in the embodiment of the present application, by constructing a multi-level coupled proxy model of components-unit bodies-overall machine, using discrete event simulation technology to perform tolerance allocation on the geometric feature parameters of components that meet the overall machine performance requirements. Through reverse tolerance allocation, it is possible to monitor whether the geometric structure parameters or unit body performance change within this range. If it exceeds, the overall machine performance may not meet the operation requirements. Also, when maintaining, it is possible to select a unit body that meets this range for replacement to ensure that the overall machine performance index requirements can be met. The present invention ensures the consistency of unit body performance through key structural feature tolerance control, and further guarantees the interchangeability of unit body functional performance, which is of great significance for promoting the agile R & D, rapid derivation, and condition-based maintenance of aero-engines. Brief description of the drawings

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0043] Figure 1 It is a flow schematic diagram of the unit body performance tolerance allocation method based on multi-level performance transfer according to an embodiment of the present invention;

[0044] Figure 2 It is a schematic diagram of the establishment process of the forward proxy model according to an embodiment of the present invention;

[0045] Figure 3 (a) is a schematic diagram of single-parameterization of triangular distribution according to an embodiment of the present invention, and (b) is a schematic diagram of single-parameterization of mixed Gaussian distribution according to an embodiment of the present invention;

[0046] Figure 4 Schematic diagram of the forward calculation workflow according to an embodiment of the present invention;

[0047] Figure 5 Schematic diagram of the reverse calculation workflow according to an embodiment of the present invention;

[0048] Figure 6 Schematic diagram of the Monte Carlo-based reverse calculation process according to an embodiment of the present invention;

[0049] Figure 7 Schematic diagram of the trade-off space according to an embodiment of the present invention. Detailed implementation manners

[0050] The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0051] The following uses specific specific examples to illustrate the implementation manners of the present application. Those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The present application can also be implemented or applied through other different specific implementation manners. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that, without conflict, the following embodiments and the features in the embodiments can be combined with each other. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts fall within the scope of protection of the present application.

[0052] It should be noted that the following describes various aspects of the embodiments within the scope of the appended claims. It should be obvious that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is illustrative only. Based on the present application, those skilled in the art should understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement the device and / or practice the method. In addition, this device can be implemented and this method can be practiced using other structures and / or functionality in addition to one or more of the aspects described herein.

[0053] It should also be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present application in a schematic manner. The diagrams only show the components related to the present application, rather than being drawn according to the number, shape, and size of the components in actual implementation. The type, quantity, and ratio of each component in actual implementation can be arbitrarily changed, and the component layout type may also be more complex.

[0054] In addition, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that the described aspects can be practiced without these specific details.

[0055] The embodiment of the present application provides a unit performance tolerance allocation method based on multi-level performance transfer. The following is a detailed description with reference to Figures 1 to 7 which specifically includes the following steps:

[0056] Step 1: Construct a forward proxy model of the whole machine including multiple unit bodies based on geometric feature parameters and whole machine performance parameters;

[0057] Step 2: In the forward proxy model of the whole machine, for each unit body, identify the key geometric feature parameters;

[0058] Step 3: For each unit body, perform dimensionality reduction of the degree of freedom on the feature distribution of each key geometric feature parameter to obtain a single parameter;

[0059] Step 4: For each unit body, input the single parameter corresponding to each key geometric feature parameter and the numerical range corresponding to each key geometric feature parameter into the forward proxy model of the whole machine for updating;

[0060] Step 5: For the whole machine target performance requirement, perform uncertainty reverse propagation based on the updated forward proxy model of the whole machine to obtain the geometric feature parameter distribution of each unit body;

[0061] Step 6: According to the obtained geometric feature parameter distribution of each unit body, use the trade-off space to obtain the tolerance allocation result.

[0062] Inverse uncertainty quantification is an effective means to determine whether the replaced unit meets the overall machine performance interchangeability. Based on the range of overall machine performance requirements, the performance tolerance range of the unit is inversely quantified, and the distribution range of further geometric feature parameters of the components is constrained. It can not only monitor whether the geometric structure parameters or the unit performance change within this range. If it exceeds, the overall machine performance may not meet the operation requirements. It can also select a unit that meets this range for replacement during maintenance to ensure that the overall machine performance index requirements can be met. Based on this feature, the embodiments of this application use geometric uncertainty factors as inputs and the unit or overall machine performance as outputs to establish a forward uncertainty propagation surrogate model; identify key geometric feature parameters through a method of feature extraction and selection by sensitivity analysis, simplify the input probability density degrees of freedom, compress the geometric feature space, and reduce the calculation dimension; based on the inverse uncertainty quantification method, calculate the selection range of geometric feature parameters under the condition of meeting the given performance requirements; finally, in a two-dimensional or three-dimensional space, graphically represent the trade-off space to visually show the selection interval of geometric feature parameters. This method helps to ensure the consistency of unit performance through the control of key structural feature tolerances, and further guarantee the interchangeability of unit functional performance, which is of great significance for promoting the agile R & D, rapid derivation, and condition-based maintenance of aero-engines.

[0063] In one embodiment, the construction of a forward surrogate model of the overall machine including multiple units based on geometric feature parameters and overall machine performance parameters includes:

[0064] For each unit, randomly select multiple samples from various geometric feature parameter degradation probabilities, perform CFD simulations on all different sample combinations, and calculate the component characteristics under the influence of geometric feature parameter degradation;

[0065] Input the component characteristics obtained by simulating each unit into a zero-dimensional overall machine performance model to obtain the overall machine performance under the combined action of various geometric feature parameter degradation modes;

[0066] Based on the obtained various overall machine performances, construct a forward surrogate model of the overall machine based on geometric feature parameters and overall machine performance parameters through the surrogate model method.

[0067] During specific implementation, consider the uncertainty of the overall machine performance caused by the combined action of degraded geometries. Randomly select multiple samples from the assumed various geometric feature degradation probabilities in a disorderly manner, perform CFD simulations on all different sample combinations, and calculate the component characteristics under the influence of geometric degradation. Substitute the component characteristics obtained by the simulation into the zero-dimensional overall machine performance model to obtain the overall machine performance under the combined action of various geometric degradation modes. Construct a forward model of geometric feature - overall machine performance parameters through the surrogate model method. The surrogate model described in this application is a model that can approximately replace the original three-dimensional high-fidelity CFD model and the performance model Y = f(X) for propagation through a certain sample set Y = f(X) is the mapping from geometric parameters to CFD and then to the whole machine performance, where Y is the whole machine performance and X is the degraded geometric parameters; In order to obtain the mapping of the whole machine performance from the geometric parameters using the proxy model, is the overall performance calculated from the proxy model, X is the degraded geometric parameter, and C is the internal coefficient of the proxy model (integrated within the machine learning model). The ultimate goal is to achieve the performance obtained by the proxy model. Approximately equal to Y. Approximate surrogate model It can save a lot of time and is therefore an essential method for uncertainty quantification in CFD. The input of the surrogate model is the uncertainty parameters of all geometric features, and the output is the performance of the entire machine.

[0068] In one embodiment, in the whole machine forward proxy model, identifying key geometric feature parameters for each unit body includes:

[0069] Performing a sensitivity analysis based on the whole machine forward surrogate model and calculating the SHAP value of each geometric characteristic parameter in each unit body;

[0070] For each unit cell, the geometric feature parameters are sorted according to the SHAP value to identify the key geometric feature parameters.

[0071] In practice, a global sensitivity analysis method is used to decompose the original model into the sum of increasing terms for different dimensional parameter combinations based on the system response variance. This allows for the quantitative calculation of the contribution of a particular uncertainty factor or combination to the total variance of the system response. Several items with significant contributions are selected for allocation in the uncertainty tolerance allocation, while the remaining features are treated as constants in the tolerance allocation.

[0072] Specifically, SHAP (SHapley Additive exPlanations) is a model interpretation method based on game theory. Its core principle comes from the Shapley value in game theory, which can quantify the marginal contribution of each participant (i.e., geometric features) to the overall benefit (i.e., model prediction). The feature contribution value of SHAP can just reflect the importance of geometric features and is suitable for sensitivity analysis. Therefore, in this embodiment, SHAP is combined with the constructed SVR agent model to perform sensitivity analysis. The overall contribution of geometric features to model prediction can be used as global sensitivity. The larger the average absolute SHAP value, the greater the impact of the geometric features on the prediction results of the model, which means that the feature is more sensitive. Therefore, the geometric feature parameters are sorted in order from large to small according to the SHAP value, and the geometric feature parameters with the highest ranking are the key geometric feature parameters.

[0073] In one embodiment, reducing the degrees of freedom of the feature distribution of each key geometric feature parameter to obtain a single parameter includes:

[0074] Representing the first four statistical moments (μx; σx; Sx; Kx) of the probability density of each key geometric feature parameter by a single parameter to complete the reduction of degrees of freedom; where μx is the mean, σx is the standard deviation, S is the skewness, and Kx is the kurtosis.

[0075] Further, reducing the degrees of freedom of the feature distribution of each key geometric feature parameter to obtain a single parameter includes:

[0076] When the feature distribution of the geometric feature parameter is a triangular distribution, defining the single parameter dx as:

[0077] dx = x ub -x lb ,

[0078] where x ub is the upper boundary of the triangular distribution, and x lb is the lower boundary of the triangular distribution.

[0079] In specific implementation, by inputting the statistical moments of the probability density function until the probability that satisfies all constraints is reached. The standard deviation can be reduced and the mean can be determined, and the same applies to other moments such as skewness (Sx) and kurtosis. Directly manipulate all input statistical moments to find various possible uncertainty reduction schemes to ensure that the input probability density is achievable. To solve this problem, a parametric function is proposed to define the shape of the probability density to reduce the input parameters for the purpose of dimensionality reduction. Among them, for the triangular distribution, referring to Figure 3 (a), defining the single parameter dx as: dx = x ub -x lb , where x ub is the upper boundary of the triangular distribution, and x lb is the lower boundary of the triangular distribution. For the influence of the adjustment of the single parameter on the probability density, for example, reshape the probability density by scaling the width relative to the nominal value x n (the mode in the triangular distribution). For the new width d' x , d' x = x' ub -x l ' b , the new parameters of the probability density are calculated from the geometric ratio and the unit area property of the triangle to obtain the corresponding width, that is, the distance between x n and the new boundaries, and has the same proportional length as shown in the following formula:

[0080]

[0081] The boundary x′ of the new triangular distribution ub 、x′ lb can be obtained from the above equation (1).

[0082] If the triangular distribution is expressed according to the normal probability density, at least three parameters are involved to fix a triangular distribution, such as the maximum value, minimum value and mode of the triangular distribution (the so-called three degrees of freedom). However, if the method of Monte Carlo sampling calculation is used and none of these three parameters are fixed, the sampling complexity will increase and the calculation process will take too long. Therefore, in this embodiment, it is selected to transform the three parameters into a parameter called dx (in the triangular distribution, dx is equal to the distance from the mode to the minimum value), and by fixing some conditions (such as the position of the mode, the ratio of the distances from the maximum and minimum values to the mode), a single parameter dx can express a triangular distribution. At this time, the degree of freedom of this triangular distribution is reduced from three dimensions to one dimension, simplifying the subsequent calculation process and reducing the calculation cost. In the above case, since the degree of freedom of the triangular distribution is reduced from three to one, it will affect all possible distribution spaces of the triangular distribution. In addition, the triangular distribution can also be parameterized by separately modifying the upper and lower bounds.

[0083] For the Gaussian distribution, referring to Figure 3 (b), for the case where the shape is between the left-skewed and right-skewed distributions, such as the symmetric distribution, the Gaussian mixture distribution f n = d x is defined as the input parameter to reshape the Gaussian mixture distribution f gm , where and are the upper and lower bounds of the given value range of μ′ x , Parameterization involves performing a linear transformation on the component functions of the following formula to obtain a new mean and a new standard deviation, which is achieved by modifying μ′ x of the position.

[0084]

[0085] where

[0086]

[0087] where μ1, μ2, σ1, σ2 are the means and variances of the original mixed Gaussian distribution, μ1', μ'2, σ1', σ'2 are the means and variances of the reshaped mixed Gaussian distribution, and α1, α2 are the weight coefficients of the reshaped mixed Gaussian distribution.

[0088] Further, the inputting of the single parameter corresponding to each key geometric feature parameter and the numerical range corresponding to each key geometric feature parameter into the whole-machine forward proxy model for updating includes:

[0089] Based on the single parameter corresponding to each key geometric feature parameter and the numerical range corresponding to each key geometric feature parameter, a mixture Gaussian distribution is constructed;

[0090] The mixture Gaussian distribution is input into the whole-machine forward proxy model for updating.

[0091] In specific implementation, for example, for the unit of the high-pressure compressor, the identified key geometric feature parameters include the tip clearance increment and the surface roughness. The dimensionality reduction of degrees of freedom is performed on these two parameters. Assuming that both parameters are Gaussian mixture distributions, the range of the surface roughness is [10, 50] μm, and the tip clearance increment is [0, 0.8] mm. A mixture Gaussian distribution based on the single parameter is constructed using the parameter dx, and then it is input into the whole-machine forward proxy model.

[0092] Further, for the whole-machine target performance requirement, based on the updated whole-machine forward proxy model, uncertainty reverse propagation is performed to obtain the geometric feature parameter distribution of each unit, including:

[0093] For the whole-machine target performance requirement, calculate the target performance standard deviation σ according to the confidence interval Ay ;

[0094] In each unit, multiple geometric feature parameters are sampled from the empirical distribution using Monte Carlo sampling to form the standard deviations of multiple geometric feature parameter distributions;

[0095] The standard deviations of multiple geometric feature parameter distributions are input into the updated whole-machine forward proxy model to calculate the standard deviation σ of the simulated distribution of the whole-machine performance Ty ;

[0096] When the difference between the standard deviation σ of the simulated distribution of the whole-machine performance Ty and the target performance standard deviation σ Ay meets the preset matching requirement, then accept the standard deviation σ of the simulated distribution of the whole-machine performance Ty ;

[0097] When the difference between the standard deviation σ of the simulated distribution of the whole-machine performance Ty and the target performance standard deviation σ Ay does not meet the preset matching requirement, then continue to adjust the multiple geometric feature parameter distributions for iterative optimization until the preset matching requirement is met;

[0098] Obtain the multiple geometric feature parameter distributions corresponding to when the preset matching requirement is met.

[0099] In specific implementation, refer to Figure 5 and 6 In each unit, use Monte Carlo sampling to extract possible geometric feature parameters from the empirical distribution to form the standard deviation of the distribution Substitute it into the forward surrogate model to calculate the standard deviation σ of the simulated distribution of the overall machine performance Ty and match it with the target performance standard deviation σ Ay If the preset matching requirement Δ = |σ Ty -σ Ay | is satisfied, then accept this σ Ty If not, continue to adjust σ x for the optimization loop. The calculation cost is the product of the number of optimization iterations and the number of runs of the sampling program in each loop. For example, each time Monte Carlo generates 10,000 samples, the maximum number of iterations is 400 for the experiment, and the results obtained from the experiment are used as the distribution of the degraded geometric parameters that meet the performance conditions.

[0100] In one embodiment, according to the obtained geometric feature parameter distribution of each unit, a tolerance allocation result is obtained by using a trade-off space, including:

[0101] Convert the obtained geometric feature parameter distribution of each unit into a trade-off space, and visually display the tolerance allocation result through a two-dimensional cross-sectional view of the trade-off space.

[0102] According to a specific implementation manner of the embodiment of the present application, the visual display of the tolerance allocation result through the two-dimensional cross-sectional view of the trade-off space includes:

[0103] Decompose the n-dimensional input space of the trade-off space into parameter pairs, where one parameter pair consists of two different geometric feature parameters. The n-dimensional input space consists of the geometric feature parameters of all the obtained units, and each parameter pair corresponds to a two-dimensional cross-section of a hypercube in the trade-off space;

[0104] Draw contour lines for each two-dimensional cross-section to obtain the m-dimensional output space of the trade-off space, and visually display the tolerance allocation structure through the m-dimensional output space; wherein, the m-dimensional output space consists of the overall machine performance parameters, and each contour line corresponds to a constraint of the deterministic or probabilistic geometric feature parameters of an overall machine performance parameter.

[0105] In specific implementation, the trade-off space can be visualized through a two-dimensional cross-sectional view of the trade-off space, and the multi-dimensional trade-off space is presented in the following way:

[0106] 1) Decompose the n-dimensional input space into parameter pairs, (x1; x2), (x3; x4)…(xn-1; xn), where x here nis the nth geometric feature parameter, and all n geometric feature parameters should be visualized in pairs. In the case where the number of parameters is odd, one parameter can be repeated. Each pair of parameters represents a plane of the hypercube, and these planes intersect at the point of the nominal value.

[0107] 2) The m-dimensional output space can be visualized through a two-dimensional cross-section of the space containing the contour lines. Each contour line represents a deterministic or probabilistic constraint in the overall machine performance parameters. Among them, the n-dimensional input space refers to the geometric parameter space, and the m-dimensional output space refers to the overall machine performance space.

[0108] Convert the degraded geometric feature parameters into a two-dimensional trade-off space, and identify the area enclosed by them by calculating and plotting the contour lines corresponding to specific overall machine performance constraints (such as temperature T6, specific fuel consumption SFC, and thrust Fn), as Figure 7 . The inner area close to the coordinate axis is the range of values that the degraded geometric parameters can take to meet the overall machine performance conditions. Designers can select values within the area according to the actual situation. The gray area exceeds the overall machine performance requirements. The dashed boundary is the optimal performance that meets the overall machine conditions, also known as the optimal boundary.

[0109] Using the trade-off space, it is possible to clearly observe the space where the geometric uncertainty can take values and the space that does not meet the requirements, providing visual support for designers to weigh the tolerance allocation results.

[0110] In one embodiment, the overall machine forward proxy model is constructed based on a machine learning method.

[0111] The following uses a specific embodiment to elaborate in detail on the unit performance tolerance allocation method based on multi-level performance transfer of the present application. Refer to Figures 1 to 7 , which specifically includes the following steps:

[0112] Taking the high-pressure compressor (HPC) unit as an example, in the HPC, due to the influence of the adverse pressure gradient on the blade surface, the geometric uncertainty of the upstream blade will be amplified in the downstream and ultimately act on the overall machine performance, resulting in the engine's performance not meeting the expected operating standards. In this example, the high-pressure compressor of an aeroengine is used as the unit, and the tolerance allocation problems of three degradation forms of its blades, namely leading-edge erosion, surface roughness, and tip clearance, are considered.

[0113] The first step: Based on the unit, construct the forward proxy model corresponding to each unit, and integrate the forward proxy models of each unit to obtain the overall machine forward proxy model.

[0114] Samples are taken through LHS, and the samples need to be shuffled to finally form 150 different DOE (Design of Experiment) combinations. The samples taken should cover the sample space as much as possible. Thus, the HPC characteristics under the combined action of multiple degradation modes can be obtained through CFD simulation and the zero-dimensional overall engine performance model. The component characteristics obtained from the simulation are brought into the zero-dimensional overall engine performance model to obtain the overall engine performance under the combined action of multiple degradation modes. A surrogate model from geometric parameters to overall engine performance is constructed based on SVR (Support Vector Regression). In the SVR surrogate model, the absolute percentage error of the HPC performance does not exceed 0.4%, the error of the overall flow rate is the largest, and the absolute percentage errors of the exhaust gas temperature and specific fuel consumption of the overall engine performance do not exceed 0.08%.

[0115] LHS is Latin Hypercube Sampling, a method for approximately random sampling from a multivariate parameter distribution. It belongs to the stratified sampling technique and is commonly used in computer experiments or Monte Carlo integration, etc.

[0116] In this step, multiple samples are randomly taken from the assumed probabilities of various geometric feature degradations, and CFD simulations are performed on all different sample combinations to calculate the component characteristics under the influence of geometric degradation. The component characteristics obtained from the simulation are brought into the zero-dimensional overall engine performance model to obtain the overall engine performance under the combined action of multiple degradation modes. Then, a forward model of geometric feature - overall engine performance parameters is constructed through the surrogate model method. The input of the surrogate model is all geometric feature uncertainty parameters, and the output is the overall engine performance.

[0117] Step 2: Identify key geometric feature parameters

[0118] Combine SHAP with the constructed SVR surrogate model for sensitivity analysis. The overall contribution of geometric feature parameters to the model prediction can be used as the global sensitivity. The larger the average absolute SHAP value, the greater the influence of the feature on the model prediction result, indicating that the feature is more sensitive. After analysis, the contribution of the tip clearance increment ΔC is the largest, while the contribution of the leading edge erosion ratio factor PF is the smallest, and for the pressure ratio of the HPC, the contributions of surface roughness and tip clearance increment are similar.

[0119] In this step, through sensitivity analysis, identify the contribution of degraded geometric parameters to their performance and operating stability, and thus focus on controlling them in the surrogate model.

[0120] Step 3: Reduce the degrees of freedom for each key geometric feature parameter

[0121] Two parameters, i.e., the tip clearance increment and the surface roughness, are selected for dimensionality reduction of degrees of freedom. Assuming that both parameters follow Gaussian mixture distributions, the range of the surface roughness is [10, 50] μm, and the tip clearance increment is [0, 0.8] mm. The Gaussian mixture distribution based on a single parameter is constructed using the parameter dx and then input into the surrogate model.

[0122] Step 4: Uncertainty backpropagation

[0123] In each element, Monte Carlo sampling is used to extract possible geometric parameters from the empirical distribution to form the standard deviation of the distribution. Substitute it into the surrogate model to calculate the simulation distribution σ of the overall machine performance. Ty , and match it with the target performance range σ Ay . If the preset matching requirement Δ = |σ Ty - σ Ay | is satisfied, then accept this σ Ty . If not, then continue to adjust σ x in the optimization loop. The computational cost is the product of the number of optimization iterations and the number of runs of the sampling program in each loop. Each Monte Carlo generates 10,000 samples, and the maximum number of iterations is 400 for experiments. The results obtained from the experiments are used as the distribution of the degraded geometric parameters that meet the performance conditions.

[0124] In this step, through the reverse calculation workflow, starting from the known overall machine performance parameters, the possible geometric degradation feature distributions that lead to these performances can be deduced to achieve the goal of tolerance allocation.

[0125] Step 5: Geometric feature tolerance allocation

[0126] The trade - off space can be visualized through the two - dimensional cross - sectional view of the trade - off space. The multi - dimensional trade - off space is presented in the following ways:

[0127] 1) Decompose the n - dimensional input space into parameter pairs, (x1; x2), (x3; x4)…(xn - 1; xn). All n parameters should be visualized in pairs. In the case where the number of parameters is odd, one parameter can be repeated. Each pair of parameters represents a plane of the hypercube. These planes intersect at the point of the nominal value.

[0128] 2) The m - dimensional output space can be visualized through the two - dimensional cross - section of the space containing contour lines. Each contour line represents a deterministic or probabilistic constraint in the overall machine performance parameters.

[0129] Transform the degraded geometry d PF and d ΔC into a two - dimensional trade - off space. Identify the region enclosed by it by calculating and plotting the contour lines corresponding to specific overall machine performance constraints (here considering the temperature T6, specific fuel consumption SFC, and thrust Fn), asFigure 7 The interior of the region is the range of values that the degraded geometric parameters can take to meet the overall machine performance conditions, and designers can select values within the region according to the actual situation. The gray part exceeds the requirements of the overall machine performance. The dashed boundary represents the optimal performance that meets the overall machine conditions, also known as the optimal boundary. By using the trade-off space, it is possible to clearly observe the space where the geometric uncertainty can take values and the space that does not meet the requirements, providing visual support for designers to weigh the results of tolerance allocation.

[0130] In the embodiment provided by the present invention, to quantify the range of the overall machine performance requirements in the reverse direction and the tolerance range of the unit body performance during the life cycle, it is necessary to first establish a forward uncertainty propagation proxy model with geometric uncertainty factors as the input and the unit body or overall machine performance as the output; identify key geometric features through the method of feature extraction and selection by sensitivity analysis, compress the feature space, and retain the key feature parameters in the scenario of condition-based maintenance application of the unit body; further simplify the input feature parameters and retain the necessary degrees of freedom to reduce the computational dimension; based on the reverse uncertainty quantification method, calculate the range of selection of feature parameters under the condition of meeting the given performance requirements; finally, in a 2D or 3D space, graphically represent the trade-off space to intuitively select the interval.

[0131] Through reverse tolerance allocation, it is possible to monitor whether the geometric structure parameters or the unit body performance change within this range. If it exceeds, the overall machine performance may not meet the operating requirements. Also, during maintenance, it is possible to select a unit body that meets this range for replacement to ensure that the overall machine performance index requirements can be met. The present invention ensures the consistency of the unit body performance through the control of the tolerances of key structural features, and further guarantees the interchangeability of the functional performance of the unit body, which is of great significance for promoting the agile research and development, rapid derivation, and condition-based maintenance of aero-engines.

[0132] As described above, this is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the technical field of this application within the technical scope disclosed by this application should be covered within the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. A method for allocating the performance tolerance of unit bodies based on multi-level performance transmission, characterized in that The method includes: Constructing a whole - machine forward proxy model containing multiple unit bodies based on geometric feature parameters and whole - machine performance parameters; In the whole - machine forward proxy model, for each unit body, identifying key geometric feature parameters; For each unit body, reducing the degree of freedom of the feature distribution of each key geometric feature parameter to obtain a single parameter; For each unit body, inputting the single parameter corresponding to each key geometric feature parameter and the numerical range corresponding to each key geometric feature parameter into the whole - machine forward proxy model for updating; For the whole - machine target performance requirements, based on the updated whole - machine forward proxy model, performing uncertainty reverse propagation to obtain the geometric feature parameter distribution of each unit body; According to the obtained geometric feature parameter distribution of each unit body, using the trade - off space to obtain the tolerance allocation result; Among them, the step of using the trade - off space to obtain the tolerance allocation result according to the obtained geometric feature parameter distribution of each unit body includes: Converting the obtained geometric feature parameter distribution of each unit body into a trade - off space, and visually displaying the tolerance allocation result through the two - dimensional cross - sectional view of the trade - off space; The step of visually displaying the tolerance allocation result through the two - dimensional cross - sectional view of the trade - off space includes: [[ID=IO]]Decomposing the n - dimensional input space of the trade - off space into parameter pairs, where one parameter pair is composed of two different geometric feature parameters, the n - dimensional input space is composed of the geometric feature parameters of all obtained unit bodies, and each parameter pair corresponds to a two - dimensional cross - section of a hypercube in the trade - off space; Drawing contour lines for each two - dimensional cross - section to obtain the m - dimensional output space of the trade - off space, and visually displaying the tolerance allocation structure through the m - dimensional output space; where the m - dimensional output space is composed of whole - machine performance parameters, and each contour line corresponds to a constraint on the geometric feature parameters with certainty or probability of a whole - machine performance parameter.

2. The method for allocating unit performance tolerances based on multi-level performance transfer according to claim 1, wherein The step of constructing a whole - machine forward proxy model containing multiple unit bodies based on geometric feature parameters and whole - machine performance parameters includes: For each unit body, randomly selecting multiple samples from various geometric feature parameter degradation probabilities, performing CFD simulations on all different sample combinations, and calculating the component characteristics under the influence of geometric feature parameter degradation; Inputting the component characteristics obtained by simulating each unit body into a zero - dimensional whole - machine performance model to obtain the whole - machine performance under the combined action of multiple geometric feature parameter degradation modes; Based on the obtained multiple whole - machine performances, constructing a whole - machine forward proxy model based on geometric feature parameters and whole - machine performance parameters through the proxy model method.

3. The unit performance tolerance allocation method based on multi-level performance transfer according to claim 1, characterized in that The step of, in the whole - machine forward proxy model, for each unit body, identifying key geometric feature parameters includes: Performing sensitivity analysis based on the whole - machine forward proxy model, and respectively calculating the SHAP value of each geometric feature parameter in each unit body; For each unit body, sorting the geometric feature parameters according to the SHAP value, and identifying the key geometric feature parameters.

4. The method for allocating the performance tolerance of a unit body based on multi-level performance transfer according to claim 1, characterized in that The step of reducing the degree of freedom of the feature distribution of each key geometric feature parameter to obtain a single parameter includes: The first four statistical moments of the probability density of each key geometric feature parameter are represented by a single parameter to complete the dimensionality reduction of degrees of freedom; where μx is the mean, σx is the standard deviation, Sx is the skewness, and Kx is the kurtosis.

5. The unit performance tolerance allocation method based on multi-level performance transfer according to claim 1, characterized in that The step of reducing the degree of freedom of the feature distribution of each key geometric feature parameter to obtain a single parameter includes: When the feature distribution of the geometric feature parameters is a triangular distribution, the single parameter dx is defined as: dx = x ub -x lb , where x ub is the upper boundary of the triangular distribution, and x lb is the lower boundary of the triangular distribution.

6. The unit performance tolerance allocation method based on multi-level performance transfer according to claim 1, characterized in that Updating the whole-machine forward proxy model by inputting the single parameter corresponding to each key geometric feature parameter and the numerical range corresponding to each key geometric feature parameter includes: Constructing a mixture Gaussian distribution based on the single parameter corresponding to each key geometric feature parameter and the numerical range corresponding to each key geometric feature parameter; Inputting the mixture Gaussian distribution into the whole-machine forward proxy model for updating.

7. The method for allocating the performance tolerance of a unit body based on multi-level performance transmission according to claim 6, wherein For the whole-machine target performance requirements, performing uncertainty backward propagation based on the updated whole-machine forward proxy model to obtain the geometric feature parameter distribution of each unit body, including: According to the target performance requirements of the whole machine, calculate the standard deviation of the target performance according to the confidence interval ; In each unit body, using Monte Carlo sampling to extract multiple geometric feature parameters from the empirical distribution to form the standard deviations of multiple geometric feature parameter distributions; Input the standard deviations of the distributions of multiple geometric feature parameters into the updated whole-machine forward proxy model to calculate the standard deviation of the simulated distribution of the whole-machine performance ; When the standard deviation of the simulated distribution of the overall machine performance and the standard deviation of the target performance meet the preset matching requirements, then accept the standard deviation of the simulated distribution of the overall machine performance ; When the standard deviation of the simulated distribution of the overall machine performance differs from the target performance standard deviation and the difference does not meet the preset matching requirements, continue to adjust the distributions of multiple geometric feature parameters for iterative optimization until the preset matching requirements are met; Obtaining multiple geometric feature parameter distributions corresponding to when the preset matching requirements are met.

8. The unit performance tolerance allocation method based on multi-level performance transfer according to any one of claims 1-7, characterized in that, The whole-machine forward proxy model is constructed based on machine learning methods.

Images