Rocket aircraft aerodynamic performance prediction method based on heaven and earth test and simulation data

Abstract

The invention discloses a rocket aircraft aerodynamic performance prediction method based on heaven and earth test and simulation data, and relates to the technical field of aerodynamic design. The method mainly comprises the following steps: obtaining a large number of aerodynamic performance parameters by using a non-viscous rapid design theory to construct a low-precision flow field training data set; performing heaven-earth correlation error correction on ground wind tunnel test data through a flight test; performing simulation error correction on a large number of aerodynamic performance parameters obtained by CFD numerical simulation through test data to obtain a high-precision training data set; and performing data fusion on the high-precision data set and the low-precision data set, and constructing a pneumatic prediction model. The invention provides the method for rapidly predicting the aerodynamic performance of the rocket aircraft, and an efficient and feasible solution is provided for rapid prediction of the aerodynamic performance.

Description

Technical Field

[0001] This invention relates to the field of aerodynamic design technology, and in particular to a method for predicting the aerodynamic performance of rockets based on ground and space test and simulation data. Background Technology

[0002] Hypersonic vehicles, with their high speed, maneuverability, and strong penetration capabilities, have become a key research focus for major aerospace powers worldwide. These characteristics place increasingly stringent demands on the evaluation of their aerodynamic performance. Achieving accurate evaluation and prediction of hypersonic aerodynamic performance is a critical challenge in the development of high-performance hypersonic vehicles. Hypersonic vehicles can be categorized into two types based on their propulsion method: glider vehicles (HGVs) and air-breathing hypersonic vehicles (HCMs). The former, being unpowered, represents one of the starting points for hypersonic vehicle research. Rocket-powered vehicles typically use rocket engines to propel themselves into space or near-space at the start of flight, then separate from the rocket and re-enter the atmosphere at high speed.

[0003] Traditional aerodynamic performance evaluation methods include theoretical analysis, numerical simulation, ground wind tunnel testing, and flight testing. Due to differences in theoretical simplification conditions, physical model accuracy, testing costs, and measurement capabilities, these methods play different roles in aircraft design, and the acquired aerodynamic data need to be cross-validated. Among these, experimental data is the most accurate, but its quantity is limited by the number of sensors and measurement capabilities, resulting in lower flow field resolution. CFD numerical simulation data is abundant and has high flow field resolution, but its accuracy is insufficient due to limitations in the number of grid cells used in the calculation, the selected turbulence model, and numerical errors.

[0004] To improve the consistency of aerodynamic data acquired by the aforementioned methods, data fusion methods are increasingly being applied to the construction of predictive models for multi-source aerodynamic data. Data fusion refers to combining multiple information sources according to a certain standard in space or time to obtain a consistent description of the object under test, thereby improving the performance of the information system. In recent years, this method has been widely used in the fusion modeling of numerical simulation and wind tunnel test data, providing an efficient and feasible solution for the rapid prediction of aircraft aerodynamic performance. Summary of the Invention

[0005] The purpose of this invention is to solve the problems in the prior art.

[0006] The technical solution adopted by this invention to solve its technical problem is: to provide a method for predicting the aerodynamic performance of rockets based on space-ground tests and simulation data, comprising the following steps:

[0007] Based on the known geometric parameters of the rocket vehicle, the relevant aerodynamic performance parameters are obtained using the rapid design theory of the inviscid Euler equation to construct a low-precision flow field training dataset.

[0008] Given the geometry of the rocket, a mesh model suitable for flow field calculation is constructed using mesh generation software.

[0009] Based on the mesh model, CFD solutions are obtained using flow field solving software.

[0010] The aerodynamic performance parameters of the rocket were obtained from the CFD solution.

[0011] By combining the aerodynamic performance parameters and geometric shape parameters of the rocket, and using the Latin hypercube method to cover a reasonable range of parameters, a numerical simulation dataset is constructed.

[0012] Conduct a limited number of flight tests on known rocket vehicles and collect flight test data;

[0013] Conduct a certain number of ground tests on known rocket vehicles and collect ground wind tunnel test data;

[0014] Error correction was performed on the ground wind tunnel test data using flight test data to obtain the corrected test dataset;

[0015] By using the corrected experimental dataset, error correction is performed on the numerical simulation dataset to obtain a high-precision flow field training dataset;

[0016] A rocket aerodynamic performance prediction model is developed by fusing low-precision and high-precision flow field training datasets.

[0017] Preferably, the step of obtaining relevant aerodynamic performance parameters from known rocket vehicle geometric parameters and constructing a low-precision flow field training dataset based on the rapid design theory of inviscid Euler equations includes the following steps:

[0018] Given the known geometric parameters of the rocket vehicle, the relevant aerodynamic performance parameters are obtained based on the rapid design theory of the inviscid Euler equation, which is expressed as:

[0019]

[0020] Where ρ is the gas density, g is the gravitational acceleration, p is the pressure, and V is the velocity field.

[0021] Using a theoretical model applicable to rocket geometry, the rocket's aerodynamic parameters are calculated analytically based on the rocket's geometric parameters. The flow field characteristics are approximated, and the Latin hypercube method is used to cover a reasonable range of parameters to generate D1 sets of parameter combinations, which serve as a low-precision flow field training dataset.

[0022] Preferably, the step of using flight test data to correct errors in ground wind tunnel test data to obtain a corrected test dataset includes the following steps:

[0023] A correction method based on physical models is adopted, including theoretical modeling correction, and a theoretical correction model for the difference between sky and ground is established based on aerodynamic theory;

[0024] Considering similar parameter analysis, we analyze the parameter differences between flight tests and wind tunnel tests, and normalize the wind tunnel test data according to similar parameters, so that the wind tunnel data and flight data can be compared and correlated based on similar parameters.

[0025] The corrected experimental data were correlated with the geometric parameters of the rocket vehicle, and the Latin hypercube method was used to cover a reasonable range of parameters to generate D2 sets of parameter combinations, which served as the corrected experimental dataset.

[0026] Preferably, the step of using the corrected experimental dataset to correct errors in the numerical simulation dataset to obtain a high-precision flow field training dataset includes the following steps:

[0027] The sampling dimension and sample size are determined based on the parameter range. A certain number of CFD calculation sample points are extracted using the Latin hypercube sampling method, and CFD calculations are performed on each sample point to obtain the corresponding pressure distribution data.

[0028] The extracted sample points and their corresponding pressure distribution data are used as input. The Kriging interpolation method based on statistical theory is used to interpolate and establish a surrogate model of the pressure distribution. The parameters of the variable to be corrected and the surrogate model are interpolated using a multinomial interpolation algorithm to generate a series of candidate sample points for subsequent optimization.

[0029] A genetic algorithm is used to optimize the objective function of the root mean square error between the test and calculated pressure distribution values, thereby obtaining the corrected values ​​for Mach number and angle of attack.

[0030] Construct a high-precision flow field training dataset to improve the consistency between CFD simulation results and experimental data.

[0031] Preferably, the geometric parameters of the rocket include the diameter of different sections of the rocket, the length of each section, and the expansion angle, and the resulting aerodynamic performance parameters include the surface pressure of the rocket.

[0032] Preferably, the calculated aerodynamic performance parameters of the rocket include the rocket lift-to-drag ratio, thrust, dynamic pressure, and normal overload.

[0033] Preferably, the flight data includes inertial altitude, Mach number, angle of attack, sideslip angle, and surface pressure at each time interval throughout the entire flight cycle.

[0034] Preferably, the ground wind tunnel test and the flight test satisfy the principle of similarity, and the test data collected by the two are consistent.

[0035] Preferably, the data fusion of the low-precision flow field training dataset and the high-precision flow field training dataset for the rocket aerodynamic performance prediction model adopts a random forest model.

[0036] Preferably, the process of fusing the low-precision flow field training dataset and the high-precision flow field training dataset to predict the aerodynamic performance of the rocket includes the following steps:

[0037] The revised experimental dataset and the low-precision flow field training dataset are combined into a merged dataset D;

[0038] Perform data preprocessing to ensure consistent input features and normalize the data.

[0039] Bootstrap sampling constructs subsets by randomly drawing m samples with replacement from the merged dataset D, and repeating this process to generate training subsets m1 = {S1, S2, ..., Sm1}, where each subset contains a mixture of samples from D1 and D2;

[0040] Single decision tree training: A decision tree Ti is trained independently for each subset Si.

[0041] For a new input xi, each decision tree outputs a predicted value yi, and the final regression result of the random forest is the average of the predictions from all trees:

[0042]

[0043] This invention offers the following advantages: Addressing the limitations of traditional aerodynamic performance evaluation methods, which are characterized by long cycles and high costs, this invention provides a method for rapidly predicting the aerodynamic performance of rockets based on inviscid rapid design theory and the fusion of simulation and experimental data. This invention better aligns with engineering practice, improves the consistency of aerodynamic data from multiple sources, and provides a new approach for evaluating the aerodynamic performance of rockets.

[0044] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments, but the present invention is not limited to the embodiments. Attached Figure Description

[0045] Figure 1 This is a diagram illustrating the method steps of an embodiment of the present invention;

[0046] Figure 2 This is a detailed flowchart of an embodiment of the present invention;

[0047] Figure 3This is a schematic diagram of a rocket vehicle according to an embodiment of the present invention;

[0048] Figure 4 This is a schematic diagram of the random forest data fusion regression model according to an embodiment of the present invention. Detailed Implementation

[0049] See Figure 1 and Figure 2 The diagram shows the method steps and detailed flowchart of the present invention, which include the following steps:

[0050] S101, based on the known geometric parameters of the rocket vehicle, uses the rapid design theory of inviscid Euler equations to obtain relevant aerodynamic performance parameters; the Euler equations are in the following form:

[0051]

[0052] Where ρ is the gas density, g is the gravitational acceleration, p is the pressure, and V is the velocity field.

[0053] Subsequently, rocket geometric parameters affecting the flow field (such as rocket slenderness ratio, nose cone radius of curvature, tail expansion angle, and wing shape parameters) are input. Using theoretical models applicable to specific rocket geometries (such as bending shock wave theory), analytical formulas (such as the Taylor-Maccoll equations) are used to calculate the rocket aerodynamic parameters, approximating the flow field characteristics. This process does not require CFD calculations. The Latin hypercube sampling (LHS) method is used to cover a reasonable range of parameters, generating D1 sets of parameter combinations to construct a low-precision flow field training dataset.

[0054] S102 is a mesh model that can be used for flow field calculations, constructed using mesh generation software based on the known geometry of the rocket.

[0055] S103, Based on the generated mesh model, perform CFD solution using flow field solving software;

[0056] S104, the aerodynamic performance parameters of the rocket were calculated from the CFD solution results;

[0057] S105: Based on the obtained aerodynamic performance parameters of the rocket vehicle, the geometric shape parameters of the rocket vehicle are correlated, and the Latin hypercube method is used to cover the reasonable range of parameters to construct a numerical simulation dataset.

[0058] S106: Conduct a small number of flight tests on known rocket vehicles and collect flight test data;

[0059] S107: Conduct a certain number of ground tests on known rocket vehicles and collect ground wind tunnel test data;

[0060] S108, using the collected flight test data, performs ground-to-ground correlation error correction on the collected ground wind tunnel test data. This correction employs a physics-based model, including theoretical modeling correction. A theoretical correction model for ground-to-ground differences is established based on aerodynamics. For rockets in hypersonic flight, rarefied gas dynamics theory can be used to correct deviations in wind tunnel data caused by rarefied gas effects. Simultaneously, similarity parameter analysis is considered, analyzing the differences in these parameters (such as Mach number and Reynolds number) between flight and wind tunnel tests. The wind tunnel test data is normalized according to similar parameters, allowing for comparison and correlation between wind tunnel and flight data based on these similar parameters. For example, in hypersonic flight, Reynolds number differences have a significant impact on aerodynamics; a Reynolds number correction factor can be used to adjust the wind tunnel data. The corrected test data is then correlated with the rocket geometric parameters from step 1. The Latin hypercube method is used to cover a reasonable range of parameters, generating parameter combinations D2 to obtain the corrected test dataset.

[0061] S109, use the corrected experimental dataset to correct simulation errors in the numerical simulation dataset. The correction process is as follows:

[0062] Based on the range of parameters to be studied (such as the range of values ​​for Mach number, angle of attack, etc.), determine the sampling dimension and sample size, use the Latin hypercube sampling method to extract a certain number of CFD calculation sample points, and perform CFD calculations on each sample point to obtain the corresponding pressure distribution data.

[0063] Using the extracted sample points and their corresponding pressure distribution data as input, a surrogate model of the pressure distribution is established using the Kriging interpolation method based on statistical theory. A multinomial interpolation algorithm is then used to interpolate the parameters of the variable to be corrected and the surrogate model, generating a series of candidate sample points. These candidate sample points will be used in the subsequent optimization process.

[0064] A genetic algorithm is used to optimize the root mean square error (objective function) between the experimental and calculated pressure distribution values, thereby obtaining the corrected values ​​for Mach number and angle of attack.

[0065] Construct a high-precision flow field training dataset to improve the consistency between CFD simulation results and experimental data;

[0066] S1010 involves fusing low-precision and high-precision flow field training datasets to establish a rocket aerodynamic performance prediction model. The specific fusion method is as follows:

[0067] Define the corrected CFD dataset D2 and the low-precision dataset D1 generated by the fast theoretical method as the merged dataset (D = D1 + D2);

[0068] Perform data preprocessing to ensure consistency of input features such as geometric parameters and incoming flow conditions, and normalize the data to avoid dimensional differences.

[0069] Bootstrap sampling constructs subsets by randomly selecting m samples with replacement from the merged dataset D (a single sample may be selected multiple times to enhance the model's focus on key data). This process is repeated to generate training subsets m1 = {S1, S2, ..., Sm1}, where each subset contains a mixture of samples from D1 and D2.

[0070] Single decision tree training: A decision tree Ti is trained independently for each subset Si.

[0071] For a new input xi, each decision tree outputs a predicted value yi, and the final regression result of the random forest is the average of the predictions from all trees:

[0072]

[0073] Specifically, with Figure 3 Taking the aircraft shown as an example, the method flow is as follows:

[0074] (1) Import the geometric parameters of different cross-section diameters, segment lengths, expansion angles, etc. of the rocket, and obtain the pressure distribution data D1 on the surface of the spacecraft based on the inviscid design theory of Euler equations;

[0075] (2) Import the rocket's geometric model. The diameter of the cylindrical computational domain is ten times the rocket's diameter, and its length is ten times the rocket's length. Construct a mesh model suitable for flow field calculations using mesh generation software. The computational domain mesh size is m.

[0076] (3) The boundary conditions are set as the pressure far field inlet total pressure p1, total temperature t1, and outlet static pressure p2. The flow field is solved by CFD using flow field solver tools to obtain the flow field numerical simulation results.

[0077] (4) The aerodynamic performance parameters of the rocket are calculated through the numerical simulation results of the flow field, including: rocket lift-to-drag ratio, thrust, dynamic pressure, normal overload, etc.

[0078] (5) Conduct n1 flight tests with a time period of t. The test data collected include the inertial altitude H, Mach number Ma, angle of attack α, sideslip angle β, engine status k (1 on, 0 off) and tail section floor pressure pt at each time node ti.

[0079] (6) Conduct n2 ground wind tunnel tests (n2 is approximately 10 times n1) with a time period of t, and collect test data of the same type as in the previous step;

[0080] (7) Conduct space-ground correlation error correction, use flight data to correct errors in ground wind tunnel data, and obtain a high-precision test dataset;

[0081] (8) Use the experimental dataset to correct the simulation error of the CFD calculation results and construct a high-precision CFD flow field training dataset D2.

[0082] (9) Use a random forest model (e.g.) Figure 4 As shown, the modified CFD flow field training dataset and the low-precision flow field dataset obtained by the fast design theory are fused. The bootstrap sampling method is used to randomly select m samples from D (D = D1 + D2) data samples to obtain m1 subsets. Each subset is trained as a decision tree. The average of the prediction results of m1 decision trees is used as the output value of the regression random forest.

[0083] As can be seen, this invention targets rocket vehicles and addresses the shortcomings of traditional aerodynamic performance evaluation methods, such as long cycles and high costs. It provides a method based on inviscid rapid design theory and the fusion of simulation and experimental data, enabling rapid prediction of rocket vehicle aerodynamic performance. This invention aligns well with engineering practice, improves the consistency of aerodynamic data from multiple sources, and provides a new approach for evaluating the aerodynamic performance of rocket vehicles.

[0084] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting the aerodynamic performance of rocket vehicles based on ground-to-ground tests and simulation data, characterized in that, Includes the following steps: Based on the known geometric parameters of the rocket vehicle, the relevant aerodynamic performance parameters are obtained using the rapid design theory of the inviscid Euler equation to construct a low-precision flow field training dataset. Given the geometry of the rocket, a mesh model suitable for flow field calculations is constructed using mesh generation software. Based on the mesh model, CFD solutions are obtained using flow field solving software. The aerodynamic performance parameters of the rocket were obtained from the CFD solution. By combining the aerodynamic performance parameters and geometric shape parameters of the rocket, and using the Latin hypercube method to cover a reasonable range of parameters, a numerical simulation dataset is constructed. Conduct a limited number of flight tests on known rocket vehicles and collect flight test data; Conduct a certain number of ground tests on known rocket vehicles and collect ground wind tunnel test data; Error correction was performed on the ground wind tunnel test data using flight test data to obtain the corrected test dataset; By using the corrected experimental dataset, the numerical simulation dataset is corrected for errors, resulting in a high-precision flow field training dataset. A rocket aerodynamic performance prediction model is developed by fusing low-precision and high-precision flow field training datasets.

2. The method for predicting the aerodynamic performance of rockets based on ground-to-ground test and simulation data according to claim 1, characterized in that, The process of obtaining relevant aerodynamic performance parameters from known rocket vehicle geometry parameters using the rapid design theory based on the inviscid Euler equations to construct a low-precision flow field training dataset includes the following steps: Given the known geometric parameters of the rocket vehicle, the relevant aerodynamic performance parameters are obtained based on the rapid design theory of the inviscid Euler equation, which is expressed as: Where ρ is the gas density, g is the gravitational acceleration, p is the pressure, and V is the velocity field. Using a theoretical model applicable to rocket geometry, the rocket's aerodynamic parameters are calculated analytically based on the rocket's geometric parameters. The flow field characteristics are approximated, and the Latin hypercube method is used to cover a reasonable range of parameters to generate D1 sets of parameter combinations, which serve as a low-precision flow field training dataset.

3. The method for predicting the aerodynamic performance of rockets based on ground-to-ground test and simulation data according to claim 1, characterized in that, The process of using flight test data to correct errors in ground wind tunnel test data to obtain a corrected test dataset includes the following steps: A correction method based on physical models is adopted, including theoretical modeling correction, and a theoretical correction model for the difference between sky and ground is established based on aerodynamic theory; Considering similar parameter analysis, we analyze the parameter differences between flight tests and wind tunnel tests, and normalize the wind tunnel test data according to similar parameters, so that the wind tunnel data and flight data can be compared and correlated based on similar parameters. The corrected experimental data were correlated with the geometric parameters of the rocket vehicle, and the Latin hypercube method was used to cover a reasonable range of parameters to generate D2 sets of parameter combinations, which served as the corrected experimental dataset.

4. The method for predicting the aerodynamic performance of rockets based on ground-to-ground test and simulation data according to claim 1, characterized in that, The process of using the corrected experimental dataset to correct errors in the numerical simulation dataset to obtain a high-precision flow field training dataset includes the following steps: The sampling dimension and sample size are determined based on the parameter range. A certain number of CFD calculation sample points are extracted using the Latin hypercube sampling method, and CFD calculations are performed on each sample point to obtain the corresponding pressure distribution data. The extracted sample points and their corresponding pressure distribution data are used as input. The Kriging interpolation method based on statistical theory is used to interpolate and establish a surrogate model of the pressure distribution. The parameters of the variable to be corrected and the surrogate model are interpolated using a multinomial interpolation algorithm to generate a series of candidate sample points for subsequent optimization. A genetic algorithm is used to optimize the objective function of the root mean square error between the test and calculated pressure distribution values, thereby obtaining the corrected values ​​for Mach number and angle of attack. Construct a high-precision flow field training dataset to improve the consistency between CFD simulation results and experimental data.

5. The method for predicting the aerodynamic performance of rockets based on ground-to-ground test and simulation data according to claim 1, characterized in that, The geometric parameters of the rocket include the diameter of different sections of the rocket, the length of each section, and the expansion angle. The resulting aerodynamic performance parameters include the surface pressure of the rocket.

6. The method for predicting the aerodynamic performance of rockets based on ground-to-ground test and simulation data according to claim 1, characterized in that, The calculated aerodynamic performance parameters of the rocket include the rocket's lift-to-drag ratio, thrust, dynamic pressure, and normal overload.

7. The method for predicting the aerodynamic performance of rockets based on ground-to-ground test and simulation data according to claim 1, characterized in that, The flight data includes inertial altitude, Mach number, angle of attack, sideslip angle, and surface pressure at each time interval throughout the entire flight cycle.

8. The method for predicting the aerodynamic performance of rockets based on ground-to-ground test and simulation data according to claim 1, characterized in that, The ground wind tunnel test and the flight test satisfy the principle of similarity, and the test data collected by the two are consistent.

9. The method for predicting the aerodynamic performance of rockets based on ground-to-ground test and simulation data according to claim 1, characterized in that, The method involves fusing low-precision and high-precision flow field training datasets to create a rocket aerodynamic performance prediction model. The data fusion employs a random forest model.

10. The method for predicting the aerodynamic performance of rockets based on ground-to-ground test and simulation data according to claim 1, characterized in that, The method for fusing low-precision and high-precision flow field training datasets to predict the aerodynamic performance of rocket vehicles includes the following steps: The revised experimental dataset and the low-precision flow field training dataset are combined into a merged dataset D; Perform data preprocessing to ensure consistent input features and normalize the data. Bootstrap sampling constructs subsets by randomly drawing m samples with replacement from the merged dataset D, and repeating this process to generate training subsets m1 = {S1, S2, ..., Sm1}, where each subset contains a mixture of samples from D1 and D2; Single decision tree training: A decision tree Ti is trained independently for each subset Si. For a new input xi, each decision tree outputs a predicted value yi, and the final regression result of the random forest is the average of the predictions from all trees:

Images