Method for formulating airworthiness conformity verification test matrix

Abstract

The invention aims to provide a method for formulating an airworthiness conformity verification test matrix, which comprises the following steps of: continuously updating a test data set by taking the prediction precision of a machine learning model on a test set as an index in a manner of combining the machine learning model and test data set iteration until the prediction precision meets the index; the training condition point set is output to serve as an aero-engine airworthiness conformity verification test matrix, and it is ensured that the airworthiness conformity verification test matrix can cover the vibration characteristics of the blade in the whole flight envelope. And obtaining the dynamic stress response of the engine blade in the whole declared flight envelope.

Description

Technical Field

[0001] The invention relates to the technical field of gas turbine engines, and in particular to a method for formulating an airworthiness compliance verification test matrix. Background Art

[0002] Airworthiness regulations CCAR33.83 requires: Civil aircraft engines must be subjected to vibration stress tests to determine whether the vibration characteristics of components that may be excited by mechanical or aerodynamic forces are acceptable within the entire declared flight envelope. At the same time, after leaving appropriate tolerances for material performance changes under various working conditions, the sum of the vibration stress and the appropriate steady-state stress must be less than the endurance limit of the relevant material.

[0003] Therefore, the vibration stress test of aircraft engines needs to cover the entire declared flight envelope, with many test conditions, long test cycles, high risks and high costs. In order to ensure the integrity and accuracy of the test results, strict requirements are imposed on the test design and result analysis methods. Testing all operating points of the flight envelope will dramatically extend the test cycle and increase the test cost, increasing the risk of engine development.

[0004] To this end, it is necessary to rationally design the test matrix to reduce the number of test conditions while obtaining the dynamic stress response of the engine blades within the entire declared flight envelope, thereby verifying the airworthiness compliance of the engine blades with respect to CCAR33.83. Summary of the invention

[0005] The object of the present invention is to provide a method for formulating an airworthiness compliance verification test matrix, which can reduce the number of test conditions while obtaining the dynamic stress response of engine blades within the entire declared flight envelope.

[0006] The method for formulating the airworthiness compliance verification test matrix to achieve the above-mentioned purpose includes the following steps:

[0007] a. Obtain the declared flight envelope of the engine;

[0008] b. Normalize parameters according to the range of all operating parameters set within the flight envelope;

[0009] c. Obtain the operating point to be tested;

[0010] d. Obtaining an initial test operating point and a set of test operating points, and placing the initial test operating point into the set of test operating points;

[0011] e. Conduct tests according to the test condition point set and establish a test data set;

[0012] f. Training turbine blade dynamic stress prediction model;

[0013] g. extracting the test operating point set into a training operating point set and a test operating point set according to any proportion, and using the test data set of the test program map corresponding to the training operating point set as a training set, and using the test data set of the test program map corresponding to the test operating point set as a test set;

[0014] h. Determine whether the prediction accuracy of the turbine blade dynamic stress prediction model on the test set meets the requirements. If so, output the blade dynamic stress prediction model and the training operating point set. If not, proceed to step i.

[0015] i. Extract the worst operating point with the worst model prediction accuracy from the test operating point set;

[0016] j. With the goal of preferentially selecting operating points near the worst operating point, select operating points in the test operating point set and add them to the test operating point set, then return to step e, and repeat steps e to h until the prediction accuracy of the turbine blade dynamic stress prediction model on the test set meets the requirements.

[0017] In one or more embodiments, the operating parameters include Mach number, pressure altitude, turbine speed, intake temperature, intake pressure, cabin pressure, ambient pressure, compressor adjustable stator blade angle, transition bleed valve angle, and fuel flow.

[0018] In one or more embodiments, the operating parameters other than the Mach number and the pressure altitude are obtained through a test program map.

[0019] In one or more embodiments, in step c, the set of operating condition points to be tested is obtained by random sampling or stratified random sampling.

[0020] In one or more embodiments, in step d:

[0021] The operating condition point to be tested located within the boundary of the flight envelope is defined as a test operating condition point, and the test operating condition point set is defined as a set of the test operating condition points;

[0022] Let the ratio of the number of training sets to the number of test sets of the blade dynamic stress prediction model be a:b, where a and b are mutually prime;

[0023] The initial test operating point is obtained by solving the following formulas (1a) to (1d):

[0024] max[min(|C i -C j |)](1a);

[0025] sC i ,Cj ∈C exp ∈C line (1b);

[0026] i≠j(1c);

[0027] crad(C exp ) = a + b (1d);

[0028] Among them, C i , C j is the initial test condition point, C exp is the set of test condition points, C line is the test condition point, C exp is the test condition point set.

[0029] In one or more embodiments, in step e, the corresponding stable operating parameters in the test program map of each operating point are extracted as input, and the maximum vibration stress parameters of the turbine blades under the corresponding parameters are extracted as output, and the test data set is established based on the input and the output.

[0030] In one or more embodiments, a suitable machine learning model is selected according to the vibration characteristics of the engine blades to train the turbine blade dynamic stress prediction model.

[0031] In one or more embodiments, the turbine blade dynamic stress prediction model is obtained by the following formula group (2a) to (2d):

[0032] minf eval (f pre ,C test )(2a);

[0033] sC train ∪C test =C exp (2b);

[0034]

[0035] crad(C train ) = m(2d);

[0036] Among them, f eval is the model evaluation function, where it is assumed that the smaller the evaluation value, the higher the model accuracy, C train is the set of training operating points, C test is the set of test operating points, m is the number of operating points used as the training operating point set, and f pre It is a prediction model for dynamic stress of turbine blades.

[0037] In one or more embodiments, the worst operating point is obtained by solving the following formulas (3a) and (3b):

[0038] maxf eval (f pre ,C worst )(3a);

[0039] sC worst ∈C test (3b);

[0040] Among them, f eval is the model evaluation function, C worst is the worst operating point, f pre It is a prediction model for dynamic stress of turbine blades.

[0041] In one or more embodiments, in step j, with the goal of preferentially selecting operating points near the worst operating point, selecting the operating points in the test operating point set is achieved by the following steps:

[0042] The distance between each operating point in the operating point set to be tested is calculated by the following formula (4):

[0043] d=min(d tresh ,|C i -C worst |),C i ∈(CC exp ) (4);

[0044] Among them, d tresh is the distance threshold. When the distance exceeds this value, it is set to d tresh , C is the set of operating points to be tested, C exp is the test condition point set;

[0045] The probability of each operating point being selected is calculated by the following formula (5):

[0046]

[0047] After completing the probability calculation of the operating point selection, the operating point is selected by the roulette wheel selection method to add to the test operating point set.

[0048] The beneficial effects of the present invention are:

[0049] The method for formulating the airworthiness compliance verification test matrix described in one or more of the aforementioned embodiments adopts a joint machine learning model and test data set iteration method, takes the prediction accuracy of the machine learning model on the test set as an indicator, continuously updates the test data set until the prediction accuracy meets the indicator, and outputs the set of training operating points as the aircraft engine airworthiness compliance verification test matrix, ensuring that the airworthiness compliance verification test matrix can cover the vibration characteristics of the blades in the entire flight envelope. Subsequently, combined with the engine airworthiness compliance verification test, the vibration stress margin of the engine blades in the flight envelope can be effectively evaluated to support engine development and airworthiness certification.

[0050] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present application. Moreover, the same reference numerals are used throughout the drawings to represent the same components. In the drawings:

[0052] Figure 1 A flowchart of some embodiments of the method for formulating the airworthiness compliance verification test matrix is ​​shown;

[0053] Figure 2 A schematic diagram showing an embodiment of an aircraft engine flight envelope;

[0054] Figure 3 A schematic diagram showing an embodiment of a commissioning procedure map is shown;

[0055] Figure 4 A distribution diagram of a set of operating condition points to be tested is shown;

[0056] Figure 5 A schematic diagram showing the distance between operating points;

[0057] Figure 6 A schematic diagram of the distribution of the test condition points after the initial test condition points are placed is shown;

[0058] Fig. 7A as well as Figure 7B Schematic diagrams of examples that meet the accuracy requirement and those that do not meet the accuracy requirement are shown respectively;

[0059] Figure 8A schematic diagram is shown after selecting an operating point to be added to a test operating point set according to an embodiment of the present application. DETAILED DESCRIPTION

[0060] The following embodiments of the technical solution of the present application are described in detail in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present application, and are therefore only used as examples, and cannot be used to limit the scope of protection of the present application.

[0061] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" in the specification and claims of this application and the above-mentioned figure descriptions and any variations thereof are intended to cover non-exclusive inclusions.

[0062] The design of existing test matrices relies on experience, and the test data can only describe the vibration stress characteristics of the aircraft engine at the test point, but cannot reflect whether it meets the airworthiness requirements in the entire flight envelope. In order to obtain a test matrix that can obtain the vibration characteristics of the aircraft engine in the entire flight envelope, according to some embodiments of the present application, a method for formulating an airworthiness compliance verification test matrix is ​​provided. Figure 1 The flowchart of some embodiments of the method for formulating the airworthiness compliance verification test matrix is ​​shown. The method for formulating the airworthiness compliance verification test matrix includes the following steps:

[0063] Step S101: Obtain the flight envelope declared by the engine, for example Figure 2 A schematic diagram of an embodiment of an aircraft engine flight envelope is shown. The aircraft envelope refers to a closed geometric figure that uses parameters such as flight speed, altitude, overload, and ambient temperature as coordinates to represent the aircraft's flight range and aircraft usage restrictions.

[0064] Step S102: normalizing parameters according to the range of all operating condition parameters set within the flight envelope.

[0065] In a specific embodiment, the operating parameters will include Mach number, pressure altitude, turbine speed, inlet temperature, inlet pressure, cabin pressure, ambient pressure, compressor adjustable stator blade angle, transition bleed valve angle and fuel flow.

[0066] Further, in a specific embodiment, the operating parameters other than the Mach number and the pressure altitude are obtained through the test program map. Specifically, the variation curves of the operating parameters other than the Mach number and the pressure altitude can be drawn in the test program map, such as Figure 3A curve diagram showing the change in parameter values ​​of turbine speed, intake temperature and cabin pressure is shown, through which the operating parameters can be obtained.

[0067] Step S103: Obtain the operating point to be tested, such as Figure 4 As shown, the gray point 1 on the aircraft engine flight envelope diagram shows a uniformly set set of test condition points C. It can be understood that Figure 4 The Mach number and pressure altitude in are both normalized parameters.

[0068] In a specific embodiment, step S103 is to obtain a set C of operating condition points to be tested by random sampling or stratified random sampling.

[0069] Step S104: obtaining an initial test operating point and a test operating point set, and placing the initial test operating point into the test operating point set.

[0070] In a specific embodiment, in step S104, the boundary of the flight envelope ( Figure 4 The solid line in the middle is the test condition point C line , the test condition point C line A set of test operating points is formed, and each operating point contains two parameters: the test Mach number and the pressure height. Assume that the ratio of the number of training sets to the number of test sets of the blade dynamic stress prediction model is a:b, where a and b are mutually prime. Then the selection of the initial test operating point is transformed into the optimization problem of the following formulas (1a) to (1d):

[0071] max[min(|C i -C j |)](1a);

[0072] sC i ,C j ∈C exp ∈C line (1b);

[0073] i≠j(1c);

[0074] crad(C exp ) = a + b (1d);

[0075] Among them, C i , C j is the initial test condition point, C exp is the set of test operating points, and the number of test operating points is a + b. By solving the optimization problem shown in formula (1a), the combination of initial test operating points can be obtained, and the shortest distance between all operating points is greater than that of other combinations.

[0076] Step S105: According to the test condition point set Cexp Conduct experiments and create experimental datasets.

[0077] In a specific embodiment, step S105 includes extracting the corresponding stable operating parameters in the test program map of each operating point, specifically including: turbine speed, intake temperature, intake pressure, cabin pressure, ambient pressure, compressor adjustable stator blade angle, transition bleed valve angle, fuel flow and other parameters. And operating point parameters: Mach number, pressure altitude. These parameters are used as input x, and the maximum vibration stress parameters of the turbine blades under the corresponding parameters are extracted as output y, and a test data set is established based on the corresponding input x and output y.

[0078] Step S106: training a turbine blade dynamic stress prediction model.

[0079] In a specific embodiment, in step S106, a suitable machine learning model is selected according to the vibration characteristics of the engine blades, such as a generalized linear model, a support vector machine, a nearest neighbor regression, a decision tree, an ensemble method, a neural network model, etc., to train the turbine blade dynamic stress prediction model f pre After completing the model training, calculate the performance of the prediction model on the test set, such as mean absolute error, mean square error, root mean square error, normalized root mean square error, determination coefficient, etc.

[0080] Step S107: According to any proportion, the test condition point set C exp Extracted as the training operating point set C train And the test condition point set C test , and the training operating point set C train The corresponding test data set of the test program map is used as the training set, and the test condition point set C test The corresponding test data set of the test program map is used as the test set.

[0081] In a specific embodiment, step S107 includes selecting m operating points as a training operating point set C train , and use its data set as the training set, and the remaining n operating points as the test operating point set C test , and use its data set as the test set. It can be understood that the number of training sets is not m, but the test data sets extracted from the test program map corresponding to the m operating points, and its number is much larger than m. Similarly, the number of test sets n is not n, but the test data sets extracted from the test program map corresponding to the n operating points, and its number is much larger than n. Among them, m and n are any coprime constants.

[0082] In a specific embodiment, the turbine blade dynamic stress prediction model is obtained by the following formula group (2a) to (2d):

[0083] minf eval (f pre ,C test )(2a);

[0084] sC train ∪C test =C exp (2b);

[0085]

[0086] crad(C train ) = m(2d);

[0087] Among them, f eval is the model evaluation function, where it is assumed that the smaller the evaluation value, the higher the model accuracy. According to formula (2a), the problem of optimizing the prediction model accuracy is solved by continuously adjusting the training operating point set C. train The combination of makes the model have the highest accuracy when predicting the test set, and the turbine blade dynamic stress model f pre .

[0088] Step S108: Determine the turbine blade dynamic stress prediction model f pre Whether the prediction accuracy on the test set meets the requirements, if so, the blade dynamic stress prediction model and the training operating point set are output; if not, proceed to step S109.

[0089] Step S109: extracting the worst operating point C with the worst model prediction accuracy in the test operating point set worst ;

[0090] In a specific embodiment, the worst operating point C worst By solving the following formulas (3a) and (3b), we can obtain:

[0091] maxf eval (f pre ,C worst )(3a);

[0092] sC worst ∈C test (3b);

[0093] Among them, it is assumed that the model evaluation function f eval The smaller the evaluation value is, the higher the model accuracy is. The worst operating point C worst The solution to the optimization problem that satisfies formula (3a) is:

[0094] Step S110: Prioritize the worst operating point C worst The nearby operating points are taken as the target, and the operating points in the test operating point set are selected to join the test operating point set C expThen, return to step S105, and repeat steps S105 to S108 until the prediction accuracy of the turbine blade dynamic stress prediction model on the test set meets the requirements.

[0095] In a specific embodiment, in step S110, the worst operating point C is selected first. worst The nearby operating point is taken as the target, and the operating point in the test operating point set is selected through the following steps:

[0096] like Figure 5 As shown, the black dot 2 is the worst operating point C worst , the gray triangle 3 and square 4 are the points in the set C of working condition points to be tested that have not been tested. The distance between each working condition point i is calculated as follows:

[0097] d=min(d tresh ,|C i -C worst |),C i ∈(CC exp ) (4);

[0098] Among them, d tresh is the distance threshold. When the distance exceeds this value, it is set to d tresh The probability of each operating point being selected is calculated as follows:

[0099]

[0100] Among them, d i Refer to formula (4) to calculate. That is, the operating point distance C worst The closer, the greater the probability of being selected. When the distance exceeds the threshold d tresh When the probability of being selected is the same, this ensures that the worst operating point C is selected first. worst Nearby operating points, improve the turbine blade dynamic stress prediction model f pre The accuracy of the worst working condition point C is uniformly selected. worst The farther point, the model is at the worst operating point C. worst And the distance from the worst operating point C worst The prediction effect of the operating point farther away can avoid the model being at the worst operating point C worst After completing the probability calculation of the operating point selection, a+b operating points are selected by roulette wheel selection to join the test operating point set C. exp Through the above method, in the test condition point set C expIn the process of increasing, the inverse of the distance between the operating points is used as the probability measurement index of the operating point being selected, and the distance threshold is set. At the same time, the advantages of preferentially selecting new operating points near the operating point with the worst prediction accuracy and selecting points farther away from the worst operating point with equal probability are taken into account. Combined with the roulette selection method, the overfitting of the model is avoided. By gradually adding working conditions based on the initial operating point set, the prediction accuracy of the blade dynamic stress is improved; while meeting the accuracy, a reasonable airworthiness compliance verification matrix is ​​formulated to minimize the number of required operating points, avoid redundancy of airworthiness compliance verification test conditions, save test costs, and reduce test cycles. Roulette selection: It belongs to the proportional selection operation. The probability distribution is regarded as a roulette wheel. The size of each slice of the roulette wheel is proportional to the probability of individual selection. The selection process is like rotating the roulette wheel. When it stops rotating, the individual corresponding to the top slice is selected. Among them, a and b in a+b are any coprime constants.

[0101] In some specific embodiments, in step S110, specifically, for the test operating point set C exp Tests are carried out on the operating points that have not been tested, and the test results are extracted to supplement the test data set. Then, a training operating point set C is randomly selected according to the ratio of m:n. train And the test condition point set C test , training turbine blade dynamic stress prediction model f pre , by adjusting the training operating point set C train The turbine blade dynamic stress prediction model f with the best prediction accuracy in the test set is obtained by combining pre , to determine whether the prediction accuracy meets the requirements. If not, select a+b operating points again according to the above method to expand the test data set until the turbine blade dynamic stress prediction model f is obtained. pre Meet the accuracy requirements and output the turbine blade dynamic stress prediction model f pre And the training set working condition, at this time the training set working condition point is the test matrix, which can be used for the airworthiness compliance verification test of the aircraft engine. Among them, a and b in a+b are any coprime constants, and m and n in m:n are any coprime constants.

[0102] The method for formulating the airworthiness compliance verification test matrix described in one or more of the aforementioned embodiments adopts a joint machine learning model and test data set iteration method, takes the prediction accuracy of the machine learning model on the test set as an indicator, continuously updates the test data set until the prediction accuracy meets the indicator, and outputs the set of training operating points as the aircraft engine airworthiness compliance verification test matrix, ensuring that the airworthiness compliance verification test matrix can cover the vibration characteristics of the blades in the entire flight envelope. Subsequently, combined with the engine airworthiness compliance verification test, the vibration stress margin of the engine blades in the flight envelope can be effectively evaluated to support engine development and airworthiness certification.

[0103] The present invention is further described below by a specific embodiment:

[0104] First, input the declared flight envelope of the engine, and normalize the operating parameters according to the range of all operating parameters set in the flight envelope. The operating parameters will include Mach number, pressure altitude, turbine speed, inlet temperature, inlet pressure, cabin pressure, ambient pressure, compressor adjustable stator blade angle, transition bleed valve angle, fuel flow, etc.

[0105] Subsequently, a set C of operating condition points to be tested is evenly set on the aircraft engine flight envelope diagram.

[0106] Subsequently, the number of training sets and the number of test sets of the turbine blade dynamic stress prediction model are set in a ratio of 3:1, and by solving the optimization problem shown in formula (1a), four initial test operating points are obtained and put into the test operating point set C. exp In, as attached Figure 6 As shown by the black dot 5 in FIG.

[0107] Then, the test condition point set C is extracted exp The test was conducted at the operating points that were not tested in the test program, and the corresponding stable turbine speed, intake temperature, intake pressure, cabin pressure, ambient pressure, compressor adjustable stator blade angle, transition bleed valve angle, fuel flow and other parameter values ​​in the test program map of each operating point were obtained, as well as the operating point parameter values: Mach number, pressure altitude. These parameters were taken as x, and the maximum vibration stress parameters of the turbine blades under the corresponding parameters were extracted as y to establish the test data set.

[0108] Then, according to the ratio of 3:1, the test condition point set C exp Extract the training operating point set C from train And the test condition point set C test , the training operating point set C train The corresponding test data set of the test program map is used as the training set, and the test condition point set C testThe corresponding test data set of the test program map is used as the test set, and the blade dynamic stress prediction model f is trained based on the decision tree model. pre , and calculate the accuracy of the model on the test set. By continuously adjusting the training operating point set C train , solve the optimization problem shown in formula (2a) and obtain the blade dynamic stress prediction model f with the best accuracy in the test set pre .

[0109] Then, the blade dynamic stress prediction model f is determined pre Whether the prediction accuracy on the test set meets the requirement, in this embodiment, if the prediction accuracy meets the 0.3 times dispersion band, it is considered to meet the accuracy requirement. Fig. 7A If the result does not meet the requirements, proceed to the next step. Figure 7B If the result shown meets the requirements, the output step is executed.

[0110] Then, solve the optimization problem shown in formula (3a) to obtain the test operating point set C test The operating point C where the model prediction accuracy is the worst worst .

[0111] According to formula (5), the set C of operating condition points not selected in the test condition points C to be tested is calculated. exp The probability of selecting the operating point is calculated, and 4 operating points are selected by roulette wheel selection to join the test operating point set C exp , as attached Figure 8 The gray dot 6 with a black outline, then returns to the Perform Test step.

[0112] When the prediction accuracy of the blade dynamic stress prediction model on the test set meets the requirements, in this embodiment, the predicted dynamic stress is within the 0.3 times dispersion band, and the blade dynamic stress prediction model f is output. pre The set of training operating points C corresponding to the model train , the operating point set is the airworthiness compliance test matrix generated by the present invention.

[0113] In the aircraft engine airworthiness compliance verification test, the test is carried out directly according to the test matrix obtained in the above steps, and the test data is extracted to train the blade dynamic stress prediction model f in the same form. pre , based on the maximum point of dynamic stress of turbine blades within the range of operating parameters predicted by the model, it is judged whether the engine meets the airworthiness requirements.

[0114] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and they should all be included in the scope of the claims and specification of the present application. In particular, as long as there is no structural conflict, the various technical features mentioned in the various embodiments can be combined in any way. The present application is not limited to the specific embodiments disclosed herein, but includes all technical solutions that fall within the scope of the claims.

Claims

1. A method for formulating an airworthiness compliance verification test matrix, characterized in that: The steps include: a. Obtain the declared flight envelope of the engine; b. Normalize parameters according to the range of all operating parameters set within the flight envelope; c. Obtain the operating point to be tested; d. Obtaining an initial test operating point and a set of test operating points, and placing the initial test operating point into the set of test operating points; e. Conduct tests according to the test condition point set and establish a test data set; f. Training turbine blade dynamic stress prediction model; g. extracting the test operating point set into a training operating point set and a test operating point set according to any proportion, and using the test data set of the test program map corresponding to the training operating point set as a training set, and using the test data set of the test program map corresponding to the test operating point set as a test set; h. Determine whether the prediction accuracy of the turbine blade dynamic stress prediction model on the test set meets the requirements. If so, output the blade dynamic stress prediction model and the training operating point set. If not, proceed to step i. i. Extract the worst operating point with the worst model prediction accuracy in the test operating point set; j. With the goal of preferentially selecting operating points near the worst operating point, select operating points in the test operating point set and add them to the test operating point set, then return to step e, and repeat steps e to h until the prediction accuracy of the turbine blade dynamic stress prediction model on the test set meets the requirements.

2. The method for formulating the airworthiness compliance verification test matrix according to claim 1, characterized in that: The operating parameters include Mach number, pressure altitude, turbine speed, intake temperature, intake pressure, cabin pressure, ambient pressure, compressor adjustable stator blade angle, transition bleed valve angle and fuel flow.

3. The method for formulating the airworthiness compliance verification test matrix according to claim 2, characterized in that: The operating parameters other than the Mach number and the pressure altitude are obtained through the test program map.

4. The method for formulating the airworthiness compliance verification test matrix according to claim 1, characterized in that: In step c, the set of operating condition points to be tested is obtained by random sampling or stratified random sampling.

5. The method for formulating the airworthiness compliance verification test matrix according to claim 1, characterized in that: In step d: The operating condition point to be tested located within the boundary of the flight envelope is defined as a test operating condition point, and the test operating condition point set is defined as a set of the test operating condition points; Let the ratio of the number of training sets to the number of test sets of the blade dynamic stress prediction model be a:b, where a and b are mutually prime; The initial test operating point is obtained by solving the following formulas (1a) to (1d): max[min(|C i -C j |)](1a); s.t.C i ,C j ∈C exp ∈C line (1b); i≠j(1c); crad(C exp )=a+b(1d); Among them, C i , C j is the initial test condition point, C exp is the set of test condition points, C line is the test condition point, C exp is the test condition point set.

6. The method for formulating the airworthiness compliance verification test matrix according to claim 1, characterized in that: In the step e, the corresponding stable operating condition parameters in the test program map of each operating point are extracted as input, the maximum vibration stress parameters of the turbine blades under the corresponding parameters are extracted as output, and the test data set is established based on the input and the output.

7. The method for formulating the airworthiness compliance verification test matrix according to claim 1, characterized in that: A suitable machine learning model is selected according to the vibration characteristics of the engine blades to train the turbine blade dynamic stress prediction model.

8. The method for formulating the airworthiness compliance verification test matrix according to claim 1, characterized in that: The turbine blade dynamic stress prediction model is obtained by the following formula group (2a) to (2d): minf eval (f pre ,C test )(2a); s.t.C train ∪C test =C exp (2b); crad(C train )=m(2d); Among them, f eval is the model evaluation function, where it is assumed that the smaller the evaluation value, the higher the model accuracy, C train is the set of training operating points, C test is the set of test operating points, m is the number of operating points used as the training operating point set, and f pre It is a prediction model for dynamic stress of turbine blades.

9. The method for formulating the airworthiness compliance verification test matrix according to claim 1, characterized in that: The worst operating point is obtained by solving the following formulas (3a) and (3b): <h2 style=";text-align:left;direction:ltr">maxf<h2 style=";text-align:left;direction:ltr"> eval <h2 style=";text-align:left;direction:ltr"> (f<h2 style=";text-align:left;direction:ltr"> pre <h2 style=";text-align:left;direction:ltr"> ,C<h2 style=";text-align:left;direction:ltr"> worst <h2 style=";text-align:left;direction:ltr"> (3a) s.t.C worst ∈C test (3b); Among them, f eval is the model evaluation function, C worst is the worst operating point, f pre It is a prediction model for dynamic stress of turbine blades.

10. The method for formulating the airworthiness compliance verification test matrix according to claim 1, characterized in that: In step j, the operating point in the test operating point set is selected with the goal of preferentially selecting the operating point near the worst operating point by the following steps: The distance between each operating point in the operating point set to be tested is calculated by the following formula (4): d=min(d tresh ,|C i -C worst |),C i ∈(C-C exp ) (4); Among them, d tresh is the distance threshold. When the distance exceeds this value, it is set to d tresh , C is the set of operating points to be tested, C exp is the test condition point set; The probability of each operating point being selected is calculated by the following formula (5): After completing the probability calculation of the operating point selection, the operating point is selected by the roulette wheel selection method to add to the test operating point set.

Images