Data enrichment method, device and electronic equipment for an engine

Abstract

Embodiments of the present application disclose a data enrichment method and device for an engine and an electronic device. After obtaining historical test data of a target engine, the historical test data is input into a pre-trained data enrichment model to obtain target test data output by the data enrichment model. The target test data includes at least one target parameter of the target engine that is not measured in historical tests. The data of the target engine that is not measured in the historical tests under corresponding historical working conditions can be quickly obtained without repeating the historical tests of the target engine, and the method is applicable to engines of any type.

Description

Technical Field

[0001] This application relates to the field of machine learning, and in particular to a data enrichment method, apparatus, and electronic device for an engine. Background Technology

[0002] In the testing of aero-engines, to gain a deeper understanding of the engine's strengths and weaknesses, it is necessary to compare test data from different engines. However, due to the different test settings and data collection methods used in different tests, it is often difficult to compare test data between engines. For example, Test 1 measured the high-pressure turbine inlet temperature of engine A, but Test 2 did not measure the high-pressure turbine inlet temperature of engine B. Therefore, it is impossible to make a horizontal comparison between engine A and engine B regarding the power distribution of the high and low pressure turbines. Another example is that Test 1 also measured the compressor after total pressure of engine A, but Test 2 only measured the compressor after static pressure of engine B. The parameters of the two are different and cannot be directly compared.

[0003] If the test data is reset by reproducing the test, the test data of the above engines are generally historical test data. The test reproduction is cumbersome, time-consuming and prone to errors. Moreover, since the engine test is relatively complex, it is difficult to reproduce the test again. Summary of the Invention

[0004] This application discloses a data enrichment method, apparatus, and electronic device for an engine, which uses a data enrichment model to obtain test data that was not measured in historical tests of the engine based on historical test data in historical tests.

[0005] According to a first aspect of the embodiments of this application, a data enrichment method for an engine is provided, the method comprising:

[0006] Obtain historical test data of the target engine; the historical test data includes historical test conditions and historical test parameters, the historical test parameters being obtained by measuring specified parameters of the target engine under historical test conditions during historical tests;

[0007] The historical test data is input into a pre-trained data enrichment model to obtain M target test data output by the data enrichment model; the M target test data include at least one target parameter of the target engine that was not measured in the historical tests.

[0008] Optionally, the data enrichment model is obtained by training multiple thermodynamic cycle models; the M target experimental data are the results of the M experimental physical quantities to be calculated in the data enrichment model;

[0009] The plurality of thermodynamic cycle models include: a first thermodynamic cycle model for determining M sensitive parameters corresponding to each experimental physical quantity; the sensitive parameters are unknown parameters used to calculate each experimental physical quantity;

[0010] And / or,

[0011] A second thermodynamic cycle model used to correct the sensitive parameters;

[0012] And / or,

[0013] The third thermodynamic cycle model is used to determine the remaining unknown parameters; the remaining unknown parameters are the parameters other than the sensitive parameters among the unknown parameters used to calculate the physical quantities of each experiment.

[0014] Optionally, the steps of the first thermodynamic cycle model in determining the M sensitive parameters corresponding to each experimental physical quantity include:

[0015] Determine the input parameters of the first thermodynamic cycle model, the input parameters including known parameters of the target engine and X unknown parameters;

[0016] The values ​​of the X unknown parameters are set to typical estimates. The known parameters and the X unknown parameters are input into the first thermodynamic cycle model to obtain the experimental physical quantity results corresponding to the M experimental physical quantities.

[0017] For each unknown parameter, the unknown parameter is modified according to a preset perturbation algorithm. The modified unknown parameter, the other unmodified unknown parameters, and the known parameters are then input into the first thermodynamic cycle model to obtain the response results of the M experimental physical quantities to the modification of the unknown parameter.

[0018] Based on the response results of each experimental physical quantity under each unknown parameter change, M sensitive parameters corresponding to each experimental physical quantity are determined from the unknown parameters; where X is greater than M.

[0019] Optionally, the step of correcting the sensitive parameter in the second thermodynamic cycle model includes:

[0020] Obtain reference data for correcting the sensitive parameters; the reference data consists of M experimental physical quantity samples obtained by measuring the target engine under specified operating conditions.

[0021] The current known parameters and X unknown parameters with estimated values ​​are input into the second thermodynamic cycle model to obtain the current experimental physical quantity results corresponding to the M experimental physical quantities;

[0022] Calculate the deviation between the current experimental physical quantity result and the baseline data;

[0023] If the deviation is greater than a preset deviation threshold, then the sensitive parameter among the X unknown parameters is corrected based on the deviation, the current estimated value of the sensitive parameter is determined to be the corrected value, and the process returns to the step of inputting the known parameter and the X unknown parameters with estimated values ​​into the second thermodynamic cycle model.

[0024] If the deviation is not greater than the preset deviation threshold, then the currently obtained sensitive parameters are determined to be the trained sensitive parameters, and the second thermal cycle model is stopped.

[0025] Optionally, the step of determining the remaining unknown parameters in the third thermodynamic cycle model includes:

[0026] Multiple pre-set training data are obtained; the training data are M experimental physical quantity training samples obtained by measuring the target engine under known operating conditions, and the operating conditions corresponding to different training data are different.

[0027] The equivalent flow area of ​​each cross section of the target engine is calculated using the physical quantity results of the specified engine finally output by the second thermodynamic cycle model and the preset Mach number.

[0028] For each training data, the equivalent flow area of ​​the cross section, the trained sensitive parameters, the remaining unknown parameters with estimated values, and the operating condition information corresponding to the training data are input into the third thermodynamic cycle model to obtain the key cross section parameters of the training data under the corresponding operating conditions.

[0029] The test physical quantity results of each training data under the corresponding working condition are calculated based on the key cross-sectional parameters. The first overall average matching degree index is obtained based on the difference between the test physical quantity results corresponding to each training data and the actual test physical quantity training samples.

[0030] Determine whether the first overall average matching degree index is less than the matching degree corresponding to the preset deviation limit. If so, determine the current remaining unknown parameters as the trained remaining unknown parameters.

[0031] If not, the remaining unknown parameters are optimized according to the preset optimization algorithm, the current estimated value of the remaining unknown parameters is determined to be the optimized value, and the process returns to the step of inputting the current known parameters and X unknown parameters with estimated values ​​into the second thermodynamic cycle model.

[0032] Optionally, the training process of the data enrichment model further includes:

[0033] Multiple pre-set verification data sets, different from the training data, are obtained. The verification data sets are M experimental physical quantity verification samples obtained by measuring the target engine under known operating conditions. Different verification data sets correspond to different operating conditions.

[0034] For each verification data, the known parameters, the trained sensitive parameters, the trained remaining unknown parameters, and the operating condition information corresponding to the verification data are input into the first thermodynamic cycle model to obtain the experimental physical quantity results of each verification data under the corresponding operating condition.

[0035] The second overall average matching degree index is obtained based on the difference between the experimental physical quantity result corresponding to each verification data and the actual experimental physical quantity verification sample. The final value is determined by judging whether the second overall average matching degree index is greater than the preset deviation limit.

[0036] If so, the currently known parameters are determined to be the trained parameters, and the first thermodynamic cycle model, which includes the currently known parameters, the trained sensitive parameters, and the trained remaining unknown parameters, is determined to be the trained data enrichment model;

[0037] If not, the known parameters are adjusted according to the preset adjustment algorithm to determine the adjusted known parameters as the current known parameters, and the process returns to the step of inputting the current known parameters and X unknown parameters with estimated values ​​into the second thermodynamic cycle model.

[0038] Optionally, the operating condition information includes at least one of the following: the engine's height above the ground, the engine's minimum Mach number, and the engine's throttle lever angle;

[0039] The reference data are selected from multiple sets of experimental physical quantity samples under different working conditions, based on the principles of minimum height, and / or minimum Mach number, and / or maximum throttle lever angle, corresponding to the specified working condition.

[0040] Optionally, the known information about the engine is the information carried in the engine's instruction manual.

[0041] According to a second aspect of the embodiments of this application, an engine data enrichment apparatus is provided, the apparatus comprising:

[0042] The acquisition unit is used to acquire historical test data of the target engine; the historical test data includes historical test conditions and historical test parameters, the historical test parameters being obtained by measuring specified parameters of the target engine under historical test conditions during historical tests;

[0043] The data enrichment unit is used to input the historical test data into a pre-trained data enrichment model to obtain M target test data output by the data enrichment model; the M target test data include at least one target parameter of the target engine that was not measured in the historical test.

[0044] According to a third aspect of the embodiments of this application, an electronic device is provided, the electronic device comprising: a memory and a processor;

[0045] The memory is used to store machine-executable instructions;

[0046] The processor is configured to read and execute machine-executable instructions stored in the memory to implement the method described above.

[0047] The technical solutions provided by the embodiments of this application may include the following beneficial effects:

[0048] As can be seen from the above technical solutions, the solution provided in this application can, after obtaining the historical test data of the target engine, input the historical test data into a pre-trained data enrichment model to obtain the target test data output by the data enrichment model. The target test data includes at least one target parameter of the target engine that was not measured in the historical test. It can quickly obtain the data of the target engine that was not measured under the historical operating conditions corresponding to the historical test without reproducing the historical test of the target engine. Moreover, this method is applicable to any type of engine.

[0049] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0050] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this specification and, together with the description, serve to explain the principles of this specification.

[0051] Figure 1 A flowchart illustrating a data enrichment method for an engine provided in this application embodiment;

[0052] Figure 2 A flowchart for determining M sensitive parameters corresponding to each experimental physical quantity using a first thermodynamic cycle model, provided in an embodiment of this application;

[0053] Figure 3 A flowchart illustrating the correction of the sensitive parameters using a second thermodynamic cycle model, as provided in this application embodiment;

[0054] Figure 4 A flowchart for determining the remaining unknown parameters using a third thermodynamic cycle model, provided as an embodiment of this application;

[0055] Figure 5 A flowchart for adjusting known parameters is provided as an embodiment of this application;

[0056] Figure 6A schematic block diagram of an engine data enrichment device provided in an embodiment of this application;

[0057] Figure 7 This is a schematic diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0058] To enable those skilled in the art to better understand the technical solutions provided in the embodiments of this application, and to make the above-mentioned objectives, features and advantages of the embodiments of this application more apparent and understandable, the method embodiments provided in the embodiments of this application are described below in conjunction with the accompanying drawings.

[0059] Please see Figure 1 , Figure 1 A flowchart of a method for enriching engine data is provided. As an example, this method can be used in electronic devices such as PCs and servers.

[0060] like Figure 1 As shown, the method includes the following steps:

[0061] Step 101: Obtain historical test data for the target engine.

[0062] In this embodiment, the historical test data of the target engine includes historical test conditions and historical test parameters. The historical test conditions refer to the condition parameters configured for the target engine in historical experiments, such as the engine's altitude above the ground, the minimum Mach number, and the throttle lever angle.

[0063] Step 102: Input the historical test data into a pre-trained data enrichment model to obtain M target test data output by the data enrichment model; the M target test data include at least one target parameter of the target engine that was not measured in the historical test.

[0064] This completes the work on... Figure 1 The description.

[0065] The method disclosed in the above embodiments allows for the input of historical test data of the target engine into a pre-trained data enrichment model after obtaining the historical test data of the target engine, thereby obtaining the target test data output by the data enrichment model. The target test data includes at least one target parameter of the target engine that was not measured in the historical test. This method can quickly obtain the data of the target engine that was not measured under the historical operating conditions corresponding to the historical test without reproducing the historical test of the target engine, and the method is applicable to any type of engine.

[0066] The training method for the data enrichment model disclosed in step 102 is described in detail below:

[0067] As an example, the data enrichment model in step 102 can be obtained by training multiple thermodynamic cycle models. The M target experimental data output in step 102 are the results of the M experimental physical quantities to be calculated in the data enrichment model.

[0068] Specifically, in this embodiment, the multiple thermodynamic cycle models used to train the data-rich model include: a first thermodynamic cycle model for determining M sensitive parameters corresponding to each experimental physical quantity; the sensitive parameters being unknown parameters used to calculate each experimental physical quantity; and / or, a second thermodynamic cycle model for correcting the sensitive parameters; and / or, a third thermodynamic cycle model for determining the remaining unknown parameters; the remaining unknown parameters being parameters other than the sensitive parameters among the unknown parameters used to calculate each experimental physical quantity.

[0069] In this embodiment, the first thermodynamic cycle model, the second thermodynamic cycle model, and the third thermodynamic cycle model use the same engine performance program, and the engine performance program in this embodiment can be any engine performance program, which is not limited in this application.

[0070] To further describe the specific process of training data-enriched models, the following section will combine... Figure 2 The process of determining the M sensitive parameters corresponding to each experimental physical quantity in the first thermodynamic cycle model described above is described in detail:

[0071] like Figure 2 As shown, it includes the following steps:

[0072] Step 201: Determine the input parameters of the first thermodynamic cycle model. The input parameters include known parameters of the target engine and X unknown parameters.

[0073] First, in order to run the first thermodynamic cycle model, this embodiment of the application needs to find the input parameters of the first thermodynamic cycle model. A portion of the input parameters can be found in the target engine's instruction manual, which contains known parameters (such as the engine's airflow rate, approximate pressure ratio, T4 temperature (i.e., turbine outlet temperature), etc.). The remaining input parameters are unknown parameters (such as the unknown parameters of various components in the engine, such as the compressor, combustion chamber, and turbine). For ease of description, in this embodiment, the X unknown parameters are denoted as {p1, p2, ..., px}, and in the following text, {p1, p2, ..., px} all represent the X unknown parameters.

[0074] Step 202: Set the values ​​of the X unknown parameters to typical estimated values, input the known parameters and the X unknown parameters into the first thermodynamic cycle model, and obtain the experimental physical quantity results corresponding to the M experimental physical quantities.

[0075] As an example, the typical estimate in step 202 refers to the value of the X unknown parameters that appears most frequently in the experiment.

[0076] It should be noted that after inputting the known parameters and X unknown parameters into the first thermodynamic cycle model in step 202, the engine physical quantity results output by the first thermodynamic cycle model based on the input parameters are generally more than the rich experimental physical quantities required in this application. Therefore, in this embodiment, after obtaining the engine physical quantity results output by the first thermodynamic cycle model, it is necessary to obtain the experimental physical quantity results corresponding to M experimental physical quantities by calculating the intersection between the output engine physical quantities and the preset experimental physical quantities.

[0077] For example, the input of the first thermodynamic cycle model is {known information, {p1, p2, ... px}}, and the final output of the engine physical quantity result is {C1, C2, ... CM, ... CQ}, while the preset experimental physical quantity is {T1, T2, ... TM}, and Q is greater than or equal to M. Then, in this embodiment, the final experimental physical quantity result is {C1, C2, ... CM}.

[0078] Step 203: For each unknown parameter, modify the unknown parameter according to the preset perturbation algorithm, and input the modified unknown parameter, the other unmodified unknown parameters, and the known parameters into the first thermodynamic cycle model to obtain the response results of the M experimental physical quantities to the modification of the unknown parameter.

[0079] As an example, the response result of each experimental physical quantity to the modification of the unknown parameter is: the ratio of the absolute value of the difference between the experimental physical quantities before and after the modification of the unknown parameter to the experimental physical quantity before the modification of the unknown parameter. Assuming the modification of an unknown parameter p1 is δ, after the modification of the unknown parameter, this embodiment inputs {known information, {p1+δ, p2, ...px}} into the first thermodynamic cycle model, and the obtained experimental physical quantity results are {C1', C2', ..., CM'}. Then, the response result of the first experimental physical quantity to the modification of the unknown parameter p1 is 1-1 = absolute value ((experimental physical quantity C1'-experimental physical quantity C1) / experimental physical quantity C1)%, and so on, recording the response results of unknown parameters p1 and experimental physical quantities one by one. In this embodiment, the corresponding results can be recorded using a table as shown in Table 1:

[0080] Table 1

[0081] Unknown parameter p1 Unknown parameter p2 Unknown parameter p3 …… Test physical quantity 1 Response 1-1 Response 1-2 Response 1-3 …… Test physical quantity 2 Response 2-1 Response 2-2 Response 2-3 …… …… …… …… …… …… Test physical quantity M Response M-1 M-2 response M-3 response ……

[0082] It should be noted that in this embodiment, the modification of each unknown parameter is based on the input in step 202, and the response results of the experimental physical quantity to the modification of the unknown parameter are based on the output in step 202.

[0083] Step 204: Based on the response results of each experimental physical quantity under the condition of each unknown parameter change, determine the M sensitive parameters corresponding to each experimental physical quantity from the unknown parameters.

[0084] As an example, any M unknown parameters are identified as sensitive parameters. For each sensitive parameter, the response results of the M experimental physical quantities to that sensitive parameter are sorted in descending order. The experimental physical quantity corresponding to the highest response result of each sensitive parameter is determined as the experimental physical quantity that matches the sensitive parameter.

[0085] If two or more sensitive parameters have the same highest response result for the same experimental physical quantity, the sensitive parameter corresponding to the highest response result is selected as the sensitive parameter corresponding to that experimental physical quantity. For other sensitive parameters that are not selected, the highest response result is excluded from the response results corresponding to the other sensitive parameters, and the experimental physical quantity corresponding to the highest response result among the remaining response results corresponding to the other sensitive parameters is determined as the experimental physical quantity matching the sensitive parameter. If there are again two or more other sensitive parameters with the highest response result among the remaining response results corresponding to the other sensitive parameters, the above steps of excluding the highest response result from the response results corresponding to the other sensitive parameters are repeated until the experimental physical quantity corresponding to each sensitive parameter is found.

[0086] By using the above method to determine the sensitive parameters, we can finally obtain the pairing relationship between the M experimental physical quantities and their corresponding sensitive parameters, as shown in Table 2.

[0087] Table 2

[0088] Pairing relationship Sensitive parameters Test physical quantity 1 Unknown parameter p1 Test physical quantity C3 2 Unknown parameter p2 Test physical quantity C1 …… …… …… M Unknown parameter pM Test physical quantity CM

[0089] This concludes the process. Figure 2 The process is shown below. (Through) Figure 2 The process shown can determine the M sensitive parameters corresponding to each experimental physical quantity through the first thermodynamic cycle model.

[0090] The following is combined Figure 3 The process of correcting the sensitive parameters in the second thermodynamic cycle model described above is described in detail below:

[0091] Step 301: Obtain reference data for correcting the sensitive parameters.

[0092] In this embodiment, the reference data consists of M experimental physical quantity samples obtained by measuring the target engine under specified operating conditions. In one embodiment, if the above operating condition information includes at least one of the following: engine height above the ground, minimum Mach number, and throttle lever angle, then the reference data can be the experimental physical quantity sample corresponding to the specified operating condition selected from multiple sets of experimental physical quantity samples under different operating conditions, according to the principle of minimum height, and / or minimum Mach number, and / or maximum throttle lever angle. In the principle of selecting the reference data, the selection priority of height, Mach number, and throttle lever angle decreases in that order. That is, if there are two sets of data, where the first set of data has the minimum height but not the minimum Mach number, and the second set of data has the minimum height but not the minimum Mach number, then the first set of data is selected as the reference data.

[0093] Step 302: Input the current known parameters and the X unknown parameters with estimated values ​​into the second thermodynamic cycle model to obtain the current experimental physical quantity results corresponding to the M experimental physical quantities.

[0094] In this embodiment, the initial estimates of the X unknown parameters are the typical design values ​​described above.

[0095] Step 303: Calculate the deviation between the current experimental physical quantity result and the baseline data.

[0096] For example, if the current test physical quantity result is {C1, C2, ..., CM} and the reference data is {T1, T2, ..., TM}, then the deviation between the current test physical quantity result and the reference data is E = {C1-T1, C2-T2, ..., CM-TM}.

[0097] Step 304: If the deviation is greater than a preset deviation threshold, then based on the deviation, correct the sensitive parameter among the X unknown parameters, determine the current estimated value of the sensitive parameter as the corrected value, and return to the step of inputting the known parameter and the X unknown parameters with estimated values ​​into the second thermodynamic cycle model.

[0098] As an example, if the deviation is greater than a preset deviation threshold, a small-range perturbation δ is applied to each of the M sensitive parameters to obtain the experimental physical quantity corresponding to each sensitive parameter after the perturbation. The new deviation E' between the experimental physical quantity corresponding to each sensitive parameter after the perturbation and the reference data is calculated. Based on the new deviation and the deviation E obtained in step 302, partial derivatives are calculated with the parameter perturbation amount to obtain the deviation correction matrix. The sensitive parameter correction amount Δ is calculated according to the formula Δparameter = inverse of the deviation correction matrix * E. The sensitive parameters are then corrected based on the sensitive parameter correction amount Δ. The Δparameter can be expressed in the form shown in Formula 1:

[0099]

[0100] Where δ1 is the parameter perturbation of the unknown parameter p1, δ M For the parameter perturbation of the unknown parameter pM, E1 is the deviation of the first experimental physical quantity from the perturbation of the unknown parameter. M It is the deviation of the Mth experimental physical quantity from the disturbance of the unknown parameter.

[0101] Step 305: If the deviation is not greater than the preset deviation threshold, then the currently obtained sensitive parameters are determined to be the trained sensitive parameters, and the second thermal cycle model is stopped.

[0102] For example, when the deviation E tends to {0, 0, ..., 0}, the currently obtained sensitive parameter is determined to be the trained sensitive parameter, and the second thermodynamic cycle model is stopped.

[0103] This concludes the process. Figure 3 The process is shown below. (Through) Figure 3 The process shown can correct M sensitive parameters.

[0104] The following is combined Figure 4 The process of correcting the sensitive parameters in the second thermodynamic cycle model described above is described in detail below:

[0105] Step 401: Obtain multiple pre-set training data sets.

[0106] In this embodiment, the training data consists of M experimental physical quantity training samples obtained by measuring the target engine under known operating conditions. Different training data correspond to different operating conditions.

[0107] Step 402: Calculate the equivalent flow area of ​​each cross section of the target engine using the final output of the specified engine physical quantity results from the second thermodynamic cycle model and the preset Mach number.

[0108] Based on the fact that the second thermodynamic cycle model and the first thermodynamic cycle model use the same engine performance program, the engine physical quantity results output by the second thermodynamic cycle model in this embodiment actually also include Q. From the Q engine physical quantity results, the parameters that need to be used when calculating the equivalent flow area of ​​the cross section are selected, such as the pressure P, temperature T, flow rate W, density ρ, specific heat ratio k, gas constant R, etc. of each cross section of the engine, as the specified engine physical quantity results.

[0109] As an example, the equivalent flow area A of each cross-section of the target engine can be calculated using the following formula 2:

[0110]

[0111] Where Ma is the Mach number.

[0112] Step 403: For each training data, the equivalent flow area of ​​the cross section, the trained sensitive parameters, the remaining unknown parameters with estimated values, and the operating condition information corresponding to the training data are input into the third thermodynamic cycle model to obtain the key cross section parameters of the training data under the corresponding operating condition.

[0113] In this embodiment, the process will be... Figure 3 The final values ​​of the sensitive parameters {p1, p2, ... pM} obtained in the process shown, the equivalent flow area A of each section obtained in step 402, the remaining unknown parameters whose current initial estimate is the typical design value, and the operating condition information corresponding to the training data are used as the inputs that have been determined for the third thermodynamic cycle model.

[0114] Step 404: Calculate the test physical quantity results of each training data under the corresponding working condition based on the key cross-sectional parameters, and obtain the first overall average matching degree index based on the difference between the test physical quantity results corresponding to each training data and the actual test physical quantity training samples.

[0115] Specifically, based on the difference between the experimental physical quantity result corresponding to each training data and the actual experimental physical quantity training sample, the matching degree index between the experimental physical quantity result corresponding to the training data and the actual experimental physical quantity training sample can be obtained by the formula (1-abs(experimental physical quantity-actual experimental physical quantity training sample) / actual experimental physical quantity training sample). Then, the matching degree index obtained for each training data is summed and averaged to obtain the first overall average matching degree index.

[0116] Step 405: Determine whether the first overall average matching degree index is less than the matching degree corresponding to the preset deviation limit. If so, determine the current remaining unknown parameters as the trained remaining unknown parameters.

[0117] Step 406: If not, optimize the remaining unknown parameters according to the preset optimization algorithm, determine the current estimated value of the remaining unknown parameters as the optimized value, and return to step 302 above.

[0118] As an example, this application uses the highest value of the first overall average matching degree index as... Figure 4 The process shown illustrates the training and optimization of the objective function. A preset optimization algorithm is used to continuously optimize the remaining unlabeled unknown parameters (denoted as {PM+1, ..., px}) until the optimized remaining unknown parameters {PM+1, ..., px} make the first overall average matching degree index less than the matching degree corresponding to a preset deviation limit (for example, setting the matching degree corresponding to the preset deviation limit to Σ(1-(1%))). The preset optimization algorithm can be a gradient-independent optimization algorithm, and this application does not impose any restrictions on it.

[0119] This concludes the process. Figure 4 The process is shown below. (Through) Figure 4 The process shown can calibrate the remaining (XM) unknown parameters among the unknown parameters.

[0120] Based on the above Figures 2-4 The process shown can calibrate the unknown parameters in the input of the first thermodynamic cycle model. However, although the known parameters used above are recorded in the engine's instruction manual, the values ​​corresponding to these known parameters can easily change due to engine manufacturing, wear and tear, etc., making the actual values ​​of the known parameters not necessarily those recorded in the manual. This can affect the calculation of the engine's experimental physical quantities. To eliminate the error caused by the known parameters in the calculation of the engine's experimental physical quantities, preferably, this application provides a method for eliminating the error caused by the known parameters in the calculation of the engine's experimental physical quantities, such as... Figure 5 As shown, it includes the following steps:

[0121] Step 501: Obtain multiple pre-set validation data that are different from the training data.

[0122] In this embodiment, the verification data consists of M experimental physical quantity verification samples obtained by measuring the target engine under known operating conditions. Different verification data correspond to different operating conditions.

[0123] Step 502: For each verification data, input the known parameters, trained sensitive parameters, trained remaining unknown parameters, and the operating condition information corresponding to the verification data into the first thermodynamic cycle model to obtain the experimental physical quantity results of each verification data under the corresponding operating condition.

[0124] Step 503: Based on the difference between the experimental physical quantity result corresponding to each verification data and the actual experimental physical quantity verification sample, obtain the second overall average matching degree index, and determine whether the second overall average matching degree index is greater than the preset deviation limit.

[0125] As an example, based on the difference between the experimental physical quantity result corresponding to each verification data point and the actual experimental physical quantity verification sample, the matching degree index between the experimental physical quantity result corresponding to the verification data and the actual experimental physical quantity verification sample can be obtained using the formula (1-abs(experimental physical quantity - actual experimental physical quantity verification sample) / actual experimental physical quantity verification sample). Then, the matching degree index obtained for each verification data point is summed and averaged to obtain the second overall average matching degree index.

[0126] Step 505: If yes, then determine that the currently known parameters are the trained parameters, and determine the first thermodynamic cycle model, which includes the currently known parameters, the trained sensitive parameters, and the trained remaining unknown parameters, as the trained data enrichment model.

[0127] Step 506: If not, adjust the known parameters according to the preset adjustment algorithm, determine the adjusted known parameters as the current known parameters, and return to step 302.

[0128] This concludes the process. Figure 5 The process is shown below. (Through) Figure 5 The process shown can be used to adjust the known parameters of the input thermodynamic cycle model to improve the accuracy of the data-rich model.

[0129] In passing Figure 5 After the process shown verifies that the accuracy of the data-rich model under operating conditions is qualified, any data-rich operating condition requirement of the target engine can be used as input to generate data-rich target test data, and the accuracy of the target test data obtained is the same as the accuracy of the verification data input into the data-rich model.

[0130] It should be noted that the baseline data, training data, and validation data in the above embodiments can be obtained in the following ways:

[0131] Test data of the target engine is collected and organized in the form of test data points. The acquired test data includes aerodynamic and thermodynamic parameters (including temperature, pressure, flow rate, etc.) of various engine sections at different heights, Mach numbers, and throttle lever angles. The acquired test data is then organized as individual engine test data points, where height, Mach number, and throttle lever angle constitute the three conditional dimensions for defining the engine test data points.

[0132] The experimental data were then divided into three categories: baseline data, training data, and validation data. A dataset consisting of N known experimental data points was used. From this dataset, one data point was selected as the baseline data according to preset principles. NT data points (e.g., NT / N < 60%) were selected as training data, and the remaining N-NT-1 data points were selected as validation data. The selection of the baseline data followed the principles of minimum height, minimum Mach number, and maximum throttle angle.

[0133] For ease of understanding, a specific embodiment will be used as an example below:

[0134] First, let's take a detailed look at the training process of a data-rich model with an example:

[0135] Before implementing this embodiment, it is necessary to obtain 10 known test data points as shown in Table 3 below. The height, Mach number, and throttle lever angle are different in these 10 test data points:

[0136] Table 3

[0137]

[0138]

[0139] Based on the selection principle of benchmark data, test data point 1, which has the lowest altitude, lowest Mach number, and highest throttle lever angle, was determined as the benchmark data. Then, the test data points other than test data point 1 were randomly divided into training data and validation data. It should be noted that when the aero engine has the lowest altitude, lowest Mach number, and highest throttle lever angle, the aero engine is generally located in the aircraft that has just started up on the ground.

[0140] It should be noted that the three parameters of intake pressure P2, intake temperature T2, and exhaust pressure PAMB in the selected test point data are known parameters that can be calculated from the data in the instruction manual. Therefore, it can be determined that the preset test physical quantities in this embodiment are air flow W2, fuel flow WF, compressor outlet temperature T3, inner inlet pressure P21, and outer bypass inlet pressure P13, and the preset number of test physical quantities is 5 (i.e., M is 5).

[0141] The input to the data-enriched model to be trained in this embodiment includes the known parameters of the engine (information recorded in the engine's manual) and the unknown parameters of the engine. The unknown parameters of the engine also include sensitive parameters associated with the experimental physical quantity and remaining unknown parameters not associated with the experimental physical quantity. In order to calibrate the known parameters and unknown parameters of the engine, this embodiment calibrates the sensitive parameters, remaining unknown parameters and known parameters in three parts respectively.

[0142] Since the sensitive parameters are related to the experimental physical quantities, and the specific volume is easy to determine and calibrate, this embodiment first determines the sensitive parameters corresponding to each experimental physical quantity through the first thermodynamic cycle model f when training the model with rich training data (wherein the engine performance program in the first thermodynamic cycle model f can be any known engine performance program, such as the commercial thermodynamic cycle program gasturb):

[0143] Step 1-1: Based on the known parameters of the engine and the input parameters of the first thermodynamic cycle model f, determine the required X unknown parameters. For example, if the first thermodynamic cycle model f has a total of 30 input parameters, and the known parameters of the engine are airflow, approximate pressure ratio, and T4 temperature, then 27 more unknown parameters are needed (i.e., X = 27). For ease of description, the embodiment will continue to be described using X = 27 as an example.

[0144] Steps 1-2: Input the known parameters and X unknown parameters into the first thermodynamic cycle model f. The X unknown parameters can be typical estimates commonly used in experiments. Obtain the Q engine physical quantity results output by the first thermodynamic cycle model f. Find the experimental physical quantity results corresponding to the 5 experimental physical quantities required in this embodiment from the Q engine physical quantity results.

[0145] For example, the input and output of the first thermodynamic cycle model f in this embodiment can be seen in Table 4 below:

[0146] Table 4

[0147]

[0148] Steps 1-3: Record the experimental physical quantities obtained in Step 1-1 as performance result 1. For each unknown parameter, modify the unknown parameter according to the preset perturbation algorithm (for example, the modification δ of the unknown parameter is taken as 5% of the reasonable variation range of the unknown parameter). Input the modified unknown parameter, the other unmodified unknown parameters, and the known parameters into the first thermodynamic cycle model to obtain the response results of the above 5 experimental physical quantities to the modification of the unknown parameter.

[0149] For example, as shown in Table 5, if the compressor efficiency p1 is modified by 0.007, based on the modified p1 and the unmodified p2-p27, five new experimental physical quantity results are obtained. These five new experimental physical quantity results are denoted as performance result 2. The calculation method for the response result of each experimental physical quantity to the modification of p1 is as follows: obtain the difference between the result of the experimental physical quantity in performance result 2 and the result in performance result 1, and determine the response result of the experimental physical quantity by the ratio of the absolute value of the difference to the result in performance result 1. For example, the response result of the compressor after-temperature T3 to the modification of p1 is: (absolute value of (experimental physical quantity C1' - experimental physical quantity C1) / experimental physical quantity C1)%.

[0150] Table 5 is shown below:

[0151] Table 5

[0152]

[0153] Based on the above calculation method, the response result of each experimental physical quantity to the change of p1 can be obtained. Furthermore, the response result of each experimental physical quantity to changes of different unknown parameters can be obtained. For example, the final response results of the experimental physical quantities to changes of different unknown parameters are shown in Table 6 below:

[0154] Table 6

[0155]

[0156] Steps 1-4: Based on the response results of the above 5 experimental physical quantities under the condition of each unknown parameter change, determine the sensitive parameters corresponding to each experimental physical quantity from the unknown parameters.

[0157] In this embodiment, the first five unknown parameters can be selected as sensitive parameters in the order of the unknown parameters. Then, the correspondence between the sensitive parameters and each experimental physical quantity is determined based on the response results shown in Table 6. Specifically, the following steps can be taken to determine the correspondence between the sensitive parameters and each experimental physical quantity: The experimental physical quantities corresponding to the sensitive parameters are determined sequentially according to their order. For each sensitive parameter, the following operation is performed: For each sensitive parameter, the response results of the five experimental physical quantities to that sensitive parameter are sorted in descending order, and the experimental physical quantity corresponding to the highest response result of each sensitive parameter is determined as the experimental physical quantity matching the sensitive parameter.

[0158] If two or more sensitive parameters have the same highest response result for the same experimental physical quantity, the sensitive parameter corresponding to the highest response result is selected as the sensitive parameter corresponding to that experimental physical quantity. For other sensitive parameters that are not selected, the highest response result is excluded from the response results corresponding to the other sensitive parameters, and the experimental physical quantity corresponding to the highest response result among the remaining response results corresponding to the other sensitive parameters is determined as the experimental physical quantity matching the sensitive parameter. If there are again two or more other sensitive parameters with the highest response result among the remaining response results corresponding to the other sensitive parameters, the above steps of excluding the highest response result from the response results corresponding to the other sensitive parameters are repeated until the experimental physical quantity corresponding to each sensitive parameter is found.

[0159] For example, in Table 6, the highest response result corresponding to the unknown parameter p1 is 1.139134, so the sensitive parameter corresponding to the experimental physical quantity C1 with a response result of 1.139134 is determined to be the unknown parameter p1; the highest response result corresponding to the unknown parameter p2 is 5.666667, so the sensitive parameter corresponding to the experimental physical quantity C2 with a response result of 5.666667 is determined to be the unknown parameter p2; the highest response result corresponding to the unknown parameter p3 is 5.666667, so the sensitive parameter corresponding to the experimental physical quantity C3 with a response result of 5.666667 is determined to be the unknown parameter p1. The sensitive parameter is the unknown parameter p3; the highest response results corresponding to the unknown parameters p4 and p5 both correspond to the experimental physical quantity C5, therefore, the higher response result 3.28209 (i.e., the highest response result corresponding to the unknown parameter p5) is selected, and the sensitive parameter corresponding to the experimental physical quantity C5 with the response result 3.28209 is determined to be the unknown parameter p5; the highest response result corresponding to the unknown parameter p4 is excluded, and the highest of the remaining response results corresponding to the unknown parameter p4, 1.459808, is determined to be the sensitive parameter corresponding to the experimental physical quantity C4, which is the unknown parameter p4. Thus, the pairing relationship between the sensitive parameters and the five experimental physical quantities is obtained, as shown in Table 7:

[0160] Table 7

[0161] Pairing relationship Sensitive parameters Test physical quantity 1 compressor efficiency compressor outlet temperature 2 Fan internal pressure ratio Import pressure 3 Fan bypass ratio Import pressure on foreign ducts 4 Convert traffic airflow 5 Combustion chamber outlet temperature Fuel flow

[0162] The following is an example of how the above-mentioned sensitive parameters were calibrated using the second thermodynamic cycle model:

[0163] Step 2-1: Input the current known parameters and 27 unknown parameters into the second thermodynamic cycle model to obtain the current experimental physical quantity results corresponding to the 5 experimental physical quantities. Among them, the initial values ​​of the 27 unknown parameters are typical design values.

[0164] The engine performance program used in the second thermodynamic cycle model is the same as that used in the first thermodynamic cycle model. The current experimental physical quantity results obtained in step 2-1 are the same as the output of the first thermodynamic cycle model f in Table 4. The current experimental physical quantity results are: experimental physical quantity C1, experimental physical quantity C2, ..., experimental physical quantity C5 = {47.335, 0.52551, 734.26, 210.0, 210.0}. Table 8 shows the input and output data of the second thermodynamic cycle model in step 2-1:

[0165] Table 8

[0166]

[0167] Step 2-2: Calculate the deviation E between the actual test physical quantity sample in the benchmark data and the current test physical quantity result.

[0168] For example, if the current experimental physical quantity result is the same as the result obtained in step 2-1, then E = {C1-T1, C2-T2, ..., C5-T5} = {734.26, 210.0, 210.0, 47.335, 0.52551} - {720.0, 188.0, 195.0, 47.30, 0.76} = {14.26, 22, 15, 0.035, -0.23449}. The sample of the actual experimental physical quantities in the baseline data can be found in Table 3.

[0169] Step 2-3: Determine whether the deviation E tends to {0, 0, ..., 0}. If so, determine the currently obtained sensitive parameter as the trained sensitive parameter and complete the calibration of the sensitive parameter. Otherwise, proceed to step 2-4.

[0170] Steps 2-4: If it is determined that the deviation E does not tend to {0, 0, ..., 0}, then a small-range perturbation is applied to each of the sensitive parameters to obtain the corresponding experimental physical quantity after the perturbation of each sensitive parameter. The new deviation E' between the experimental physical quantity after the perturbation of each sensitive parameter and the baseline data is calculated. Based on the new deviation E' and the deviation E obtained in step 302, partial derivatives are calculated with the parameter perturbation amount to obtain the deviation correction matrix.

[0171] For example, if each of the sensitive parameters is perturbed by a small range, the new deviation E' between the perturbation amount of each sensitive parameter and the experimental physical quantity and the reference data is shown in Table 9. Table 9 records the deviation of each experimental physical quantity when the sensitive parameter is perturbed:

[0172] Table 9

[0173]

[0174] Then, according to Formula 1, the correction amount Δ for the sensitive parameter can be calculated as follows:

[0175]

[0176] Step 2-5: Correct the sensitive parameters based on the calculated sensitive parameter correction amount. Based on the 27 unknown parameters and known parameters including the corrected sensitive parameters, return to step 2-1 for iterative loop until the sensitive parameters are calibrated in step 2-3, and stop running the second thermodynamic cycle model.

[0177] In actual operation, this embodiment requires multiple iterations, which will not be elaborated here. Based on the data in the tables given in the above embodiment, the final values ​​of the sensitivity parameters p1-p5 obtained through experiments are: 0.8089, 2.6857, 2.7857, 68.4490, and 1910.4471.

[0178] This completes the calibration of the aforementioned sensitive parameters using the second thermodynamic cycle model.

[0179] The following is an example of how to calibrate the remaining unknown parameters using the third thermodynamic cycle model:

[0180] Based on the fact that the sensitive parameters are unknown parameters p1-p5 determined through the first thermodynamic cycle model f, the remaining unknown parameters in this embodiment are p6-p27.

[0181] Step 3-1: Input the specified engine physical quantity results (including pressure P, temperature T, flow rate W, density ρ, specific heat ratio k, and gas constant R of each section of the engine) and the preset Mach number into Formula 2 to obtain the equivalent flow area A of each section of the target engine.

[0182] Formula 2 is:

[0183] Based on the data in Tables 3-9 above, the equivalent flow area A of each cross section of the target engine obtained in this embodiment is shown in Table 10:

[0184]

[0185] Step 3-2: For each training data point (4 in total) in Table 3, input the equivalent flow area of ​​the cross section, the trained sensitive parameters, the remaining unknown parameters, and the corresponding operating condition information of the training data into the third thermodynamic cycle model to obtain the key cross section parameters of the training data under the corresponding operating condition. The initial values ​​of the remaining unknown parameters are typical estimates.

[0186] Step 3-2: Based on the difference between the experimental physical quantity result and the actual experimental physical quantity training sample corresponding to each training data, the matching degree index between the experimental physical quantity result and the actual experimental physical quantity training sample corresponding to the training data is obtained by formula (1-abs(experimental physical quantity-actual experimental physical quantity training sample) / actual experimental physical quantity training sample). Then, the matching degree index obtained for each training data is summed and averaged to obtain the first overall average matching degree index.

[0187] For example, as shown in Table 11, the experimental physical quantity results corresponding to the training data are the data in bold in Table 11. The matching degree indexes obtained for each group of training data are 4.985345755, 4.979603572, 4.945173405, and 4.936047801. The first overall average matching degree index = (4.985345755 + 4.979603572 + 4.945173405 + 4.936047801) / 20 = 0.99230.

[0188] Table 11

[0189]

[0190]

[0191] Step 3-3: Determine whether the first overall average matching degree index is less than the matching degree corresponding to the preset deviation limit. If so, determine the current remaining unknown parameters as the trained remaining unknown parameters; otherwise, proceed to step 3-4.

[0192] Step 3-4: If the first overall average matching degree index is not less than the matching degree corresponding to the preset deviation limit, then optimize the remaining unknown parameters according to the preset optimization algorithm, determine the current estimated value of the remaining unknown parameters as the optimized value, and return to execute the above steps 2-1-3-4.

[0193] Based on the data from the above embodiments, the final values ​​of the remaining unknown parameters {P6,…p27} obtained through optimization in this embodiment are shown in Table 12:

[0194] Table 12 Final values ​​of remaining unknown parameters after optimization

[0195]

[0196]

[0197] This completes the calibration of the remaining unknown parameters using the third thermodynamic cycle model.

[0198] The following example illustrates the calibration process for the aforementioned known parameters (intake pressure P2, intake temperature T2, exhaust pressure PAMB):

[0199] Step 4-1: For each of the 5 verification data in Table 3, input the known parameters, the trained sensitive parameters, the trained remaining unknown parameters, and the operating condition information corresponding to the verification data into the first thermodynamic cycle model to obtain the experimental physical quantity results of each verification data under the corresponding operating condition.

[0200] Step 4-2: Obtain the second overall average matching degree index based on the difference between the experimental physical quantity results corresponding to each verification data and the actual experimental physical quantity verification sample.

[0201] For example, as shown in Table 13, the experimental physical quantity results corresponding to the verification data are the data in bold in Table 13. The matching degree indexes obtained for each group of training data are 4.980107188, 4.97834059, 4.987903019, 4.976149294, and 4.968851418. The second overall average matching degree index = (4.980107188 + 4.97834059 + 4.987903019 + 4.976149294 + 4.968851418) / 25 = 24.89135151 / 25 = 0.99565.

[0202] Table 13

[0203]

[0204]

[0205] Step 4-3: Determine whether the second overall average matching degree index calculated in step 4-4 is greater than the preset deviation limit. If so, determine that the current known parameters are the trained parameters, and determine the first thermal cycle model, which includes the current known parameters, the trained sensitive parameters, and the trained remaining unknown parameters, as the trained data enrichment model. Otherwise, execute step 4-1.

[0206] Step 4-4: If the second overall average matching degree index is not greater than the preset deviation limit, then the known parameters are adjusted according to the preset adjustment algorithm, and the adjusted known parameters are determined as the current known parameters. Then, return to step 2-1 to step 4-4.

[0207] At this point, the calibration of the known parameters is complete, and a well-trained, data-rich model is obtained.

[0208] In this embodiment, for a well-trained, data-rich model, any operating condition information of the target transmitter can be input to obtain the target test data corresponding to the test physical quantities of the target engine under any operating condition.

[0209] The methods provided in the embodiments of this application have been described above. The apparatus provided in the embodiments of this application is described below:

[0210] See Figure 6 , Figure 6 This application provides a data enrichment device for an engine. As one embodiment, this device can be used in electronic devices such as PCs and servers. The device includes:

[0211] The acquisition unit 601 is used to acquire historical test data of the target engine; the historical test data includes historical test conditions and historical test parameters, the historical test parameters being obtained by measuring specified parameters of the target engine under historical test conditions during historical tests.

[0212] The data enrichment unit 602 is used to input the historical test data into a pre-trained data enrichment model to obtain M target test data output by the data enrichment model; the M target test data include at least one target parameter of the target engine that was not measured in the historical test.

[0213] Optionally, the data enrichment model is obtained by training multiple thermodynamic cycle models; the M target experimental data are the results of the M experimental physical quantities to be calculated in the data enrichment model;

[0214] The plurality of thermodynamic cycle models include: a first thermodynamic cycle model for determining M sensitive parameters corresponding to each experimental physical quantity; the sensitive parameters are unknown parameters used to calculate each experimental physical quantity;

[0215] And / or,

[0216] A second thermodynamic cycle model used to correct the sensitive parameters;

[0217] And / or,

[0218] The third thermodynamic cycle model is used to determine the remaining unknown parameters; the remaining unknown parameters are the parameters other than the sensitive parameters among the unknown parameters used to calculate the physical quantities of each experiment.

[0219] Optionally, the steps of the first thermodynamic cycle model in determining the M sensitive parameters corresponding to each experimental physical quantity include:

[0220] Determine the input parameters of the first thermodynamic cycle model, the input parameters including known parameters of the target engine and X unknown parameters;

[0221] The values ​​of the X unknown parameters are set to typical estimates. The known parameters and the X unknown parameters are input into the first thermodynamic cycle model to obtain the experimental physical quantity results corresponding to the M experimental physical quantities.

[0222] For each unknown parameter, the unknown parameter is modified according to a preset perturbation algorithm. The modified unknown parameter, the other unmodified unknown parameters, and the known parameters are then input into the first thermodynamic cycle model to obtain the response results of the M experimental physical quantities to the modification of the unknown parameter.

[0223] Based on the response results of each experimental physical quantity under each unknown parameter change, M sensitive parameters corresponding to each experimental physical quantity are determined from the unknown parameters; where X is greater than M.

[0224] Optionally, the step of correcting the sensitive parameter in the second thermodynamic cycle model includes:

[0225] Obtain reference data for correcting the sensitive parameters; the reference data consists of M experimental physical quantity samples obtained by measuring the target engine under specified operating conditions.

[0226] The current known parameters and X unknown parameters with estimated values ​​are input into the second thermodynamic cycle model to obtain the current experimental physical quantity results corresponding to the M experimental physical quantities;

[0227] Calculate the deviation between the current experimental physical quantity result and the baseline data;

[0228] If the deviation is greater than a preset deviation threshold, then the sensitive parameter among the X unknown parameters is corrected based on the deviation, the current estimated value of the sensitive parameter is determined to be the corrected value, and the process returns to the step of inputting the known parameter and the X unknown parameters with estimated values ​​into the second thermodynamic cycle model.

[0229] If the deviation is not greater than the preset deviation threshold, then the currently obtained sensitive parameters are determined to be the trained sensitive parameters, and the second thermal cycle model is stopped.

[0230] Optionally, the step of determining the remaining unknown parameters in the third thermodynamic cycle model includes:

[0231] Multiple pre-set training data are obtained; the training data are M experimental physical quantity training samples obtained by measuring the target engine under known operating conditions, and the operating conditions corresponding to different training data are different.

[0232] The equivalent flow area of ​​each cross section of the target engine is calculated using the physical quantity results of the specified engine finally output by the second thermodynamic cycle model and the preset Mach number.

[0233] For each training data, the equivalent flow area of ​​the cross section, the trained sensitive parameters, the remaining unknown parameters with estimated values, and the operating condition information corresponding to the training data are input into the third thermodynamic cycle model to obtain the key cross section parameters of the training data under the corresponding operating conditions.

[0234] The test physical quantity results of each training data under the corresponding working condition are calculated based on the key cross-sectional parameters. The first overall average matching degree index is obtained based on the difference between the test physical quantity results corresponding to each training data and the actual test physical quantity training samples.

[0235] Determine whether the first overall average matching degree index is less than the matching degree corresponding to the preset deviation limit. If so, determine the current remaining unknown parameters as the trained remaining unknown parameters.

[0236] If not, the remaining unknown parameters are optimized according to the preset optimization algorithm, the current estimated value of the remaining unknown parameters is determined to be the optimized value, and the process returns to the step of inputting the current known parameters and X unknown parameters with estimated values ​​into the second thermodynamic cycle model.

[0237] Optionally, the training process of the data enrichment model further includes:

[0238] Multiple pre-set verification data sets, different from the training data, are obtained. The verification data sets are M experimental physical quantity verification samples obtained by measuring the target engine under known operating conditions. Different verification data sets correspond to different operating conditions.

[0239] For each verification data, the known parameters, the trained sensitive parameters, the trained remaining unknown parameters, and the operating condition information corresponding to the verification data are input into the first thermodynamic cycle model to obtain the experimental physical quantity results of each verification data under the corresponding operating condition.

[0240] The second overall average matching degree index is obtained based on the difference between the experimental physical quantity result corresponding to each verification data and the actual experimental physical quantity verification sample. The final value is determined by judging whether the second overall average matching degree index is greater than the preset deviation limit.

[0241] If so, the currently known parameters are determined to be the trained parameters, and the first thermodynamic cycle model, which includes the currently known parameters, the trained sensitive parameters, and the trained remaining unknown parameters, is determined to be the trained data enrichment model;

[0242] If not, the known parameters are adjusted according to the preset adjustment algorithm to determine the adjusted known parameters as the current known parameters, and the process returns to the step of inputting the current known parameters and X unknown parameters with estimated values ​​into the second thermodynamic cycle model.

[0243] Optionally, the operating condition information includes at least one of the following: the engine's height above the ground, the engine's minimum Mach number, and the engine's throttle lever angle;

[0244] The reference data are selected from multiple sets of experimental physical quantity samples under different working conditions, based on the principles of minimum height, and / or minimum Mach number, and / or maximum throttle lever angle, corresponding to the specified working condition.

[0245] Optionally, the known information about the engine is the information carried in the engine's instruction manual.

[0246] This completes the work on... Figure 6 Description of the device shown.

[0247] Correspondingly, embodiments of this application also provide a hardware structure diagram of an electronic device, specifically as follows: Figure 7 As shown, this electronic device can be a device for implementing the above-described engine data enrichment method. For example... Figure 7 As shown, the hardware architecture includes a processor and memory.

[0248] The memory is used to store machine-executable instructions;

[0249] The processor is configured to read and execute machine-executable instructions stored in the memory to implement a method embodiment of the corresponding engine data enrichment method as shown above.

[0250] As one embodiment, the memory can be any electronic, magnetic, optical, or other physical storage device that can contain or store information such as executable instructions, data, etc. For example, the memory can be volatile memory, non-volatile memory, or similar storage media. Specifically, the memory can be RAM (Random Access Memory), flash memory, storage drives (such as hard disk drives), solid-state drives, any type of storage disk (such as optical discs, DVDs, etc.), or similar storage media, or combinations thereof.

[0251] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for enriching engine data, characterized in that, The method includes: Obtain historical test data of the target engine; the historical test data includes historical test conditions and historical test parameters, the historical test parameters being obtained by measuring specified parameters of the target engine under historical test conditions during historical tests; The historical test data is input into a pre-trained data enrichment model to obtain M target test data output by the data enrichment model; the M target test data include at least one target parameter of the target engine that was not measured in the historical test; the data enrichment model is obtained by training multiple thermodynamic cycle models; the M target test data are the results of M test physical quantities to be calculated in the data enrichment model; The plurality of thermodynamic cycle models include: a first thermodynamic cycle model for determining M sensitive parameters corresponding to each experimental physical quantity; the sensitive parameters are unknown parameters used to calculate each experimental physical quantity; And / or, A second thermodynamic cycle model used to correct the sensitive parameters; And / or, A third thermodynamic cycle model is used to determine the remaining unknown parameters; the remaining unknown parameters are the parameters other than the sensitive parameters among the unknown parameters used to calculate the physical quantities of each experiment. The steps for correcting the sensitive parameters in the second thermodynamic cycle model include: Obtain reference data for correcting the sensitive parameters; the reference data consists of M experimental physical quantity samples obtained by measuring the target engine under specified operating conditions. The current known parameters and X unknown parameters with estimated values ​​are input into the second thermodynamic cycle model to obtain the current experimental physical quantity results corresponding to the M experimental physical quantities; Calculate the deviation between the current experimental physical quantity result and the baseline data; If the deviation is greater than a preset deviation threshold, then the sensitive parameter among the X unknown parameters is corrected based on the deviation, the current estimated value of the sensitive parameter is determined to be the corrected value, and the process returns to the step of inputting the known parameter and the X unknown parameters with estimated values ​​into the second thermodynamic cycle model. If the deviation is not greater than the preset deviation threshold, then the currently obtained sensitive parameter is determined to be the trained sensitive parameter, and the second thermal cycle model is stopped. The steps for determining the remaining unknown parameters in the third thermodynamic cycle model include: Multiple pre-set training data are obtained; the training data are M experimental physical quantity training samples obtained by measuring the target engine under known operating conditions, and the operating conditions corresponding to different training data are different. The equivalent flow area of ​​each cross section of the target engine is calculated using the physical quantity results of the specified engine finally output by the second thermodynamic cycle model and the preset Mach number. For each training data, the equivalent flow area of ​​the cross section, the trained sensitive parameters, the remaining unknown parameters with estimated values, and the operating condition information corresponding to the training data are input into the third thermodynamic cycle model to obtain the key cross section parameters of the training data under the corresponding operating conditions. The test physical quantity results of each training data under the corresponding working condition are calculated based on the key cross-sectional parameters. The first overall average matching degree index is obtained based on the difference between the test physical quantity results corresponding to each training data and the actual test physical quantity training samples. Determine whether the first overall average matching degree index is less than the matching degree corresponding to the preset deviation limit. If so, determine the current remaining unknown parameters as the trained remaining unknown parameters. If not, the remaining unknown parameters are optimized according to the preset optimization algorithm, the current estimated value of the remaining unknown parameters is determined to be the optimized value, and the process returns to the step of inputting the current known parameters and X unknown parameters with estimated values ​​into the second thermodynamic cycle model.

2. The method according to claim 1, characterized in that, The steps for determining the M sensitive parameters corresponding to each experimental physical quantity in the first thermodynamic cycle model include: Determine the input parameters of the first thermodynamic cycle model, the input parameters including known parameters of the target engine and X unknown parameters; The values ​​of the X unknown parameters are set to typical estimates. The known parameters and the X unknown parameters are input into the first thermodynamic cycle model to obtain the experimental physical quantity results corresponding to the M experimental physical quantities. For each unknown parameter, the unknown parameter is modified according to a preset perturbation algorithm. The modified unknown parameter, the other unmodified unknown parameters, and the known parameters are then input into the first thermodynamic cycle model to obtain the response results of the M experimental physical quantities to the modification of the unknown parameter. Based on the response results of each experimental physical quantity under each unknown parameter change, M sensitive parameters corresponding to each experimental physical quantity are determined from the unknown parameters; where X is greater than M.

3. The method according to claim 2, characterized in that, The training process of the data enrichment model also includes: Multiple pre-set verification data sets, different from the training data, are obtained. The verification data sets are M experimental physical quantity verification samples obtained by measuring the target engine under known operating conditions. Different verification data sets correspond to different operating conditions. For each verification data, the known parameters, the trained sensitive parameters, the trained remaining unknown parameters, and the operating condition information corresponding to the verification data are input into the first thermodynamic cycle model to obtain the experimental physical quantity results of each verification data under the corresponding operating condition. The second overall average matching degree index is obtained based on the difference between the experimental physical quantity result corresponding to each verification data and the actual experimental physical quantity verification sample. The final value is determined by judging whether the second overall average matching degree index is greater than the preset deviation limit. If so, the currently known parameters are determined to be the trained parameters, and the first thermodynamic cycle model, which includes the currently known parameters, the trained sensitive parameters, and the trained remaining unknown parameters, is determined to be the trained data enrichment model; If not, the known parameters are adjusted according to the preset adjustment algorithm to determine the adjusted known parameters as the current known parameters, and the process returns to the step of inputting the current known parameters and X unknown parameters with estimated values ​​into the second thermodynamic cycle model.

4. The method according to any one of claims 1-3, characterized in that, The operating condition information includes at least one of the following: the engine's height above the ground, the engine's minimum Mach number, and the engine's throttle lever angle. The reference data are selected from multiple sets of experimental physical quantity samples under different working conditions, based on the principles of minimum height, and / or minimum Mach number, and / or maximum throttle lever angle, corresponding to the specified working condition.

5. The method according to any one of claims 1 or 2, characterized in that, The known information about the engine is the information contained in the engine's instruction manual.

6. A data enrichment device for an engine, characterized in that, The device includes: The acquisition unit is used to acquire historical test data of the target engine; the historical test data includes historical test conditions and historical test parameters, the historical test parameters being obtained by measuring specified parameters of the target engine under historical test conditions during historical tests; The data enrichment unit is used to input the historical test data into a pre-trained data enrichment model to obtain M target test data output by the data enrichment model; the M target test data include at least one target parameter of the target engine that was not measured in the historical test; the data enrichment model is obtained by training multiple thermodynamic cycle models; the M target test data are the results of M test physical quantities to be calculated in the data enrichment model; The plurality of thermodynamic cycle models include: a first thermodynamic cycle model for determining M sensitive parameters corresponding to each experimental physical quantity; the sensitive parameters are unknown parameters used to calculate each experimental physical quantity; And / or, A second thermodynamic cycle model used to correct the sensitive parameters; And / or, A third thermodynamic cycle model is used to determine the remaining unknown parameters; the remaining unknown parameters are the parameters other than the sensitive parameters among the unknown parameters used to calculate the physical quantities of each experiment. The steps for correcting the sensitive parameters in the second thermodynamic cycle model include: Obtain reference data for correcting the sensitive parameters; the reference data consists of M experimental physical quantity samples obtained by measuring the target engine under specified operating conditions. The current known parameters and X unknown parameters with estimated values ​​are input into the second thermodynamic cycle model to obtain the current experimental physical quantity results corresponding to the M experimental physical quantities; Calculate the deviation between the current experimental physical quantity result and the baseline data; If the deviation is greater than a preset deviation threshold, then the sensitive parameter among the X unknown parameters is corrected based on the deviation, the current estimated value of the sensitive parameter is determined to be the corrected value, and the process returns to the step of inputting the known parameter and the X unknown parameters with estimated values ​​into the second thermodynamic cycle model. If the deviation is not greater than the preset deviation threshold, then the currently obtained sensitive parameter is determined to be the trained sensitive parameter, and the second thermal cycle model is stopped. The steps for determining the remaining unknown parameters in the third thermodynamic cycle model include: Multiple pre-set training data are obtained; the training data are M experimental physical quantity training samples obtained by measuring the target engine under known operating conditions, and the operating conditions corresponding to different training data are different. The equivalent flow area of ​​each cross section of the target engine is calculated using the physical quantity results of the specified engine finally output by the second thermodynamic cycle model and the preset Mach number. For each training data, the equivalent flow area of ​​the cross section, the trained sensitive parameters, the remaining unknown parameters with estimated values, and the operating condition information corresponding to the training data are input into the third thermodynamic cycle model to obtain the key cross section parameters of the training data under the corresponding operating conditions. The test physical quantity results of each training data under the corresponding working condition are calculated based on the key cross-sectional parameters. The first overall average matching degree index is obtained based on the difference between the test physical quantity results corresponding to each training data and the actual test physical quantity training samples. Determine whether the first overall average matching degree index is less than the matching degree corresponding to the preset deviation limit. If so, determine the current remaining unknown parameters as the trained remaining unknown parameters. If not, the remaining unknown parameters are optimized according to the preset optimization algorithm, the current estimated value of the remaining unknown parameters is determined to be the optimized value, and the process returns to the step of inputting the current known parameters and X unknown parameters with estimated values ​​into the second thermodynamic cycle model.

7. An electronic device, characterized in that, The electronic device includes: a memory and a processor; The memory is used to store machine-executable instructions; The processor is configured to read and execute machine-executable instructions stored in the memory to implement the method as described in any one of claims 1-5.

Images