An inlet passage performance parameter prediction method and model combining a characteristic line method

Abstract

The application provides an inlet performance parameter prediction method and model combined with a characteristic line method, and belongs to the field of hypersonic vehicle aerodynamic design. The method solves the problems of low sample utilization rate, long modeling period and limited prediction accuracy of an existing proxy model. The inlet performance parameter prediction method combined with the characteristic line method comprises the following steps: determining a plurality of geometric design parameters representing an inlet shape; constructing an initial sample set; screening the initial sample set according to a shock adhesion criterion to obtain an effective sample set meeting a critical working condition; carrying out viscous numerical simulation on each inlet configuration in the effective sample set to complete sample labeling; dividing the labeled effective sample set into a training set and a verification set to form a training data set; training a neural network by using the training data set, checking the model accuracy by using the verification set, and obtaining an inlet performance prediction proxy model. The method is mainly used for inlet performance parameter prediction.

Description

Technical Field

[0001] This invention belongs to the field of hypersonic vehicle aerodynamic design, and in particular relates to a method and model for predicting inlet performance parameters using the method of characteristics. Background Technology

[0002] Supersonic and hypersonic air-breathing propulsion systems are widely used in space transportation, spacecraft, and other fields. As the core component of the propulsion system, the air intake is responsible for capturing high-Mach number incoming airflow and compressing it to provide the downstream combustion chamber with a high-pressure airflow with sufficient flow and excellent flow field quality. As the Mach number of aircraft continues to increase, the industry has placed higher demands on the total pressure recovery capability, airflow pressurization capability, and operational stability of the air intake.

[0003] Curved compression inlets rely on curved shock waves to achieve stepwise airflow compression, offering advantages such as short axial dimensions and high compression efficiency, making them the mainstream configuration for hypersonic inlets. Currently, there are two main technical approaches for obtaining inlet aerodynamic performance: one is inviscid flow field calculation based on the method of characteristics to solve the Euler equations. This method is computationally efficient, but based on the inviscid assumption, it cannot account for viscous effects such as shock waves, boundary layer interference, and flow separation, resulting in significant deviations from actual operating conditions. Post-design corrections based on human experience are necessary. The other approach is viscous numerical simulation by solving the Navier-Stokes equations. This method can fully simulate viscous flow details with high computational accuracy, but simulations of a single configuration are time-consuming and resource-intensive.

[0004] Currently, the industry commonly uses data-driven surrogate models to quickly predict inlet performance. The conventional approach involves sampling the inlet's geometric parameters across the entire domain and then performing viscous numerical simulations on all sampled configurations to create a training dataset. However, curved compression inlets are subject to shock wave lip-fitting critical operating constraints, causing many sampled configurations to fall into subcritical or supercritical invalid operating conditions, rendering them unsuitable for engineering optimization. Invalid samples not only significantly increase the annotation costs of numerical simulations but also create dataset redundancy, reducing the training efficiency and generalization ability of neural networks. Existing surrogate model construction schemes do not perform pre-sample screening for inlet critical operating conditions, resulting in low sample utilization, long modeling cycles, and limited prediction accuracy, making it difficult to meet the engineering requirements for rapid iterative design of hypersonic inlets. Summary of the Invention

[0005] In view of this, the present invention aims to propose an inlet performance parameter prediction method that combines the method of characteristics, in order to solve the problems of low sample utilization, long modeling cycle and limited prediction accuracy of existing surrogate models.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: According to a first aspect of the present invention, a method for predicting inlet performance parameters combining the method of characteristics is provided, comprising the following steps:

[0007] The curved surface compression inlet is geometrically parameterized to determine several geometric design parameters that characterize the shape of the inlet.

[0008] Sampling is performed within the range of values ​​for each geometric design parameter to generate multiple sets of intake duct geometric configurations, forming an initial sample set;

[0009] The inviscid flow field of each inlet configuration in the initial sample set was calculated using the method of characteristics to obtain the corresponding shock wave position. The initial sample set was then screened according to the shock wave lip judgment criterion to obtain an effective sample set that satisfies the critical operating condition.

[0010] Viscous numerical simulations were performed on each inlet configuration in the effective sample set to extract the corresponding aerodynamic performance parameters and complete the sample labeling.

[0011] The labeled valid sample set is divided into a training set and a validation set to form a training dataset.

[0012] The neural network is trained using the training dataset, and the model accuracy is verified using the validation set to obtain the inlet performance prediction surrogate model.

[0013] Furthermore, the geometric design parameters include a first type of parameters that control the shape of the outer compression surface of the intake duct, and a second type of parameters that control the shape of the intake duct lip cover.

[0014] Furthermore, the intake duct geometry is generated by using Latin hypercube sampling to generate multiple intake duct geometry configurations.

[0015] Furthermore, the shock wave position is obtained by solving the supersonic inviscid flow control equation using the swirling characteristic line method, and calculating the curved shock wave position corresponding to the inlet.

[0016] Furthermore, the shock wave lip determination criterion is as follows: when the ordinate of the end of the curved shock wave falls into a preset value range, the current intake configuration is determined to be a valid sample of the critical operating condition.

[0017] Furthermore, the viscosity numerical simulation is performed by solving the Reynolds-averaged Navier-Stokes equations and combining them with a turbulence model; the aerodynamic performance parameters include the total pressure recovery coefficient and the static pressure rise coefficient.

[0018] Furthermore, the effective sample set is divided into a training set and a validation set in an 8:2 ratio; the neural network adopts a fully connected deep neural network and uses the Adam optimizer and mean squared error loss function to complete the network training.

[0019] According to a second aspect of the present invention, an inlet performance parameter prediction model combining the method of characteristics is provided, comprising:

[0020] The parameterization module is used to perform geometric parameterization on the curved surface compression intake, and to determine several geometric design parameters that characterize the shape of the intake.

[0021] The initial sample set construction module samples within the range of values ​​for each geometric design parameter to generate multiple sets of intake duct geometric configurations, thus forming the initial sample set.

[0022] The effective sample set acquisition module is used to perform inviscid flow field calculations on each inlet configuration in the initial sample set using the method of characteristics, obtain the corresponding shock wave position, and screen the initial sample set according to the shock wave lip judgment criterion to obtain an effective sample set that meets the critical operating conditions.

[0023] The sample labeling module is used to perform viscous numerical simulation on each inlet configuration in the effective sample set, extract the corresponding aerodynamic performance parameters, and complete the sample labeling.

[0024] The partitioning module divides the labeled valid sample set into a training set and a validation set to form the training dataset.

[0025] The training and verification module trains a neural network using the training dataset and verifies the model accuracy using a verification set to obtain an intake duct performance prediction proxy model.

[0026] According to a third aspect of the present invention, an electronic device is provided, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement an intake performance parameter prediction method combining the characteristic line method as described above.

[0027] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for predicting inlet performance parameters using the method of characteristic lines.

[0028] Compared with the prior art, the beneficial effects of the present invention are:

[0029] 1. This method utilizes the fast computation speed of the method of characteristics to quickly screen a large number of configurations, retaining only the critical working condition samples that are valid in engineering. Then, numerical simulation annotations that take into account the viscosity effect are carried out on the valid samples. This not only avoids the problem of insufficient accuracy of the non-viscous algorithm, but also greatly reduces the computational scale of the viscosity simulation, achieving a balance between efficiency and accuracy.

[0030] 2. This method relies on the working mechanism of the air intake and uses shock wave contact lip as the criterion to select samples. It only creates labeled data for effective critical samples, which greatly reduces the workload of sample labeling and modeling cycle. At the same time, the entire dataset consists of samples that meet the constraints of actual work, making the data more representative and improving the training effect of the subsequent neural network from the data source level.

[0031] 3. This method uses numerical simulation results that incorporate the complete viscous effect as the model training labels. Unlike prediction models trained solely on inviscid flow field data, the trained surrogate model accurately reproduces the performance changes caused by viscous flow and can realistically characterize the nonlinear mapping relationship between the inlet geometry and the total pressure recovery coefficient and static pressure rise coefficient. After model training, the validation set loss converges to... The overall prediction accuracy of the model meets the requirements for engineering use.

[0032] 4. In traditional design processes, performance parameters for each inlet configuration must be obtained through time-consuming viscous simulation, resulting in extremely low efficiency in batch configuration evaluation and parameter pattern analysis. The surrogate model trained by this invention can directly and quickly output aerodynamic performance results based on the input geometric parameters, rapidly completing performance prediction and completely replacing the traditional repetitive viscous simulation work. This significantly accelerates the progress of forward design, parameter iteration, and batch configuration selection for hypersonic curved surface compression inlets.

[0033] 5. The overall technical approach of this method is not limited by a single flight Mach number. It can complete sample screening, data labeling and model training for typical hypersonic incoming flow conditions such as 4.5, 5.0, 5.5 and 6.0 respectively, and build a dedicated proxy model for the corresponding conditions. It can meet the design and evaluation needs of curved compression inlets under different flight speeds, and has a wide range of applications and strong portability. Attached Figure Description

[0034] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0035] Figure 1 This is a schematic diagram of the curved air intake configuration with far-field pressure boundary and pressure outlet of the present invention.

[0036] Figure 2 This is a schematic diagram illustrating the three types of computational nodes solved by the spiral characteristic line method of this invention;

[0037] Figure 3 This is a schematic diagram illustrating the calculation of the bending shock wave location using the characteristic line method combined with the recursive method of the present invention.

[0038] Figure 4 This is a contour map of the outer compression segment calculated using the characteristic line method of this invention;

[0039] Figure 5 This is a diagram showing the pressure distribution on the outer compression section wall under different configurations and incoming flow conditions according to the present invention;

[0040] Figure 6 This is a comparison diagram of the bending shock wave using the characteristic line method of this invention and CFD numerical schlieren.

[0041] Figure 7 This is a diagram showing the absolute error between the abscissa of the final shock wave position and the abscissa of the fairing lip in this invention.

[0042] Figure 8 This is a diagram illustrating a critical state calculation using the characteristic line method of this invention.

[0043] Figure 9 This is the loss function used in the training process of the proxy model of this invention;

[0044] Figure 10 To visualize the relative error distribution of all samples in the validation set;

[0045] Figure 11 For different lip mask parameters The following is a comparison chart of the prediction performance of the numerical solution and the surrogate model;

[0046] Figure 12 Density gradient cloud maps of the inner compression section under different lip mask parameters at Mach numbers of 4.5 and 6.0;

[0047] Figure 13 A curve showing the positive design characteristics of the lip cover profile when the compression surface is fixed.

[0048] Figure 14 The static pressure distribution diagram of the lip cover wall for the six sets of calculation examples in Table 3;

[0049] Figure 15 The density gradient cloud map of the inner compression section for the six sets of examples in Table 3;

[0050] Figure 16 This is a structural block diagram of an inlet performance parameter prediction model combining the method of characteristics according to the present invention.

[0051] Figure 17 This is a block diagram of the electronic device. Detailed Implementation

[0052] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0053] The following description, with reference to the accompanying drawings, illustrates an inlet performance parameter prediction method, model, electronic device, and storage medium based on the method of characteristics, according to embodiments of the present invention. This addresses the problems mentioned in the background section regarding the low sample utilization, long modeling cycle, and limited prediction accuracy of existing surrogate models.

[0054] Specifically, a method for predicting inlet performance parameters using the method of characteristics includes the following steps:

[0055] The curved surface compression inlet is geometrically parameterized to determine several geometric design parameters that characterize the shape of the inlet.

[0056] Sampling is performed within the range of values ​​for each geometric design parameter to generate multiple sets of intake duct geometric configurations, forming an initial sample set;

[0057] The inviscid flow field of each inlet configuration in the initial sample set was calculated using the method of characteristics to obtain the corresponding shock wave position. The initial sample set was then screened according to the shock wave lip judgment criterion to obtain an effective sample set that satisfies the critical operating condition.

[0058] Viscous numerical simulations were performed on each inlet configuration in the effective sample set to extract the corresponding aerodynamic performance parameters and complete the sample labeling.

[0059] The labeled valid sample set is divided into a training set and a validation set to form a training dataset.

[0060] The neural network is trained using the training dataset, and the model accuracy is verified using the validation set to obtain the inlet performance prediction surrogate model.

[0061] Specifically, the surrogate model uses the intake duct's geometric configuration parameters as input, and parametric modeling is the foundation for achieving quantitative representation of these geometric parameters. This embodiment studies a curved intake duct with external and internal compression structures. The overall intake duct configuration and computational domain boundaries are set as follows: Figure 1 As shown, the intake duct compression surface consists of a tangent straight line segment and two parabolic curves, which are tangent to the straight line at positions (x0, y0) and (x1, y1), respectively. The compression surface can be expressed as:

[0062]

[0063] In the formula, δ0, L, h, and H are all constants, taken as δ0 = 100 mm, L = 900 mm, h = 350 mm, and H = 400 mm, respectively. Substituting x1 = δ0 + L into the equation of the outer compression section, the corresponding ordinate y1 = (tanθ1 - tanθ0) / δ0 can be obtained. 2+2tanθ0·δ0; The slope of the tangent at (x1, y1) is taken to its maximum value. The angle between this tangent and the horizontal direction is θ1. To avoid ill-conditioned curves, the maximum value of θ1 is defined as θ. 1,max =arctan[2(h-tanθ0·δ0) / L-tanθ0], and introduce parameter λ1 to adjust θ1, satisfying θ1=λ1θ 1,max In summary, the compression surface is controlled by two parameters, θ0 and λ1, with values ​​ranging from [6°, 12°] to [0.6, 0.9].

[0064] Lip position Keeping it fixed, the inner wall of the lip cover is also constructed using a parabolic shape, the expression of which is:

[0065]

[0066] Through parameters control The slope of the tangent at that point. The value range is [0, 0.8]. The above... , To control the first type of geometric parameters of the external compression surface, The second type of geometric parameters for controlling the shape of the lip cover.

[0067] The working aerodynamic data of the above-mentioned air intake were obtained through computational fluid dynamics numerical simulation. In this embodiment, within the range of values ​​of each geometric parameter, Latin hypercube sampling was used to extract samples, and finally 1000 sets of air intake configurations with different combinations of geometric parameters were generated as the initial sample set.

[0068] This embodiment obtains inlet aerodynamic data through computational fluid dynamics numerical simulation, and the solution model is equipped with... The two-dimensional Reynolds-averaged Navier-Stokes equations for the turbulence model were used. The computational domain mesh was generated using ICEM CFD, and the flow field solution was implemented using ANSYS Fluent software. Specific simulation settings are shown in Table 1. The static pressure rise coefficient π and the total pressure recovery coefficient σ were selected as indices for evaluating the aerodynamic performance of the inlet.

[0069] Hypersonic incoming flow forms a curved shock wave after passing through the external compression section of the inlet. Different streamlines generate entropy gradients after crossing the shock wave, which in turn create vorticity in the flow field. This type of flow belongs to vortex supersonic flow. For this flow pattern, this embodiment uses the vortex characteristic line method for solution. The corresponding characteristic lines and compatibility equations are as follows:

[0070]

[0071] In the formula, Γ ± Let θ be the slope of the characteristic line, θ represent the airflow deflection angle, and μ be the Mach angle, satisfying the calculation formula μ=arcsin(1 / Ma); ρ is the fluid density, and V is the airflow velocity. Therefore, C... ± The characteristic line is the Mach line. For steady adiabatic inviscid supersonic flow, the flow along the same streamline strictly follows the isentropic law, and the compatibility equation established along the streamline is expressed as:

[0072] In the formula, Γ0 is the streamline slope, and c is the speed of sound. In summary, the swirling characteristic line method includes two Mach characteristic lines and one streamline, combined with four sets of compatible equations, which can solve for the flow field parameters at various points downstream of a curved shock wave. The points to be solved are mainly divided into three categories: wall points, points inside the flow field, and shock wave points. The solution forms for these three types of points are as follows: Figure 2 As shown.

[0073] The specific solution process is as follows:

[0074] When solving for the wall point, it is necessary to know an upstream wall point and an interior point of the flow field, such as... Figure 2 As shown in (a). Passing through interior point 1. The characteristic line (Mach line) intersects the C0 characteristic line (streamline) passing through the wall point 3, and the intersection point is the wall point 4 to be solved. The difference equation with respect to the characteristic line C0 can be expressed as:

[0075]

[0076]

[0077] The subscripts in the formula correspond to Figure 2 At each feature point, γ is the specific heat ratio, and its value is set to 1.4. First, the initial slope of the line connecting 1-4 is set to the slope of the Mach line at point 1. The position of the intersection point 4 can be determined by simultaneously solving the equations of the straight line and the wall profile. In the interval δ0≤x<δ0+L, the wall profile equation can be rearranged into y=a using equation (1). wall x 2 +b walll x+c wall In the form of, where the wall coefficients are respectively The airflow deflection angle θ4 can be taken as the slope of the wall tangent at point 4, i.e., θ4 = arctan(2a). wall x 4 +b wallThen, the pressure p4 is obtained from equation (3), and the density ρ4, sound speed c4 and Mach number Ma4 are calculated using equation (4). After calculating the residuals of x4, y4, θ4, p4, ρ4 and Ma4, Γ1 is updated using the newly solved Mach angle μ4. The iteration loop ends when the residuals meet the convergence tolerance requirements.

[0078] Solving for point 4 inside the unknown flow field requires two known upstream points, such as... Figure 2 As shown in (b), the difference equation for the C+ characteristic line is:

[0079]

[0080] Point 3 is the intersection of the streamline passing through flow point 4 and the line connecting points 1 and 2. In solving for point 4, in addition to estimating the slopes of the Mach lines 1-4 and 2-4, the slope of the streamline 3-4 also needs to be determined. By solving for the intersection point, the coordinates of points 4 and 3 can be determined. Then, the flow field parameters of point 3 can be obtained by interpolating the parameters of points 1 and 2. By simultaneously solving the compatibility equations along the streamline and Mach lines, the various physical parameters of point 4 can be solved. After completing the residual calculation and slope update, the next iteration cycle can begin.

[0081] When solving for the shock point, in addition to the characteristic lines and compatibility equations, the flow relationships before and after the shock wave are also required. Solving for the flow relationships before and after an oblique shock wave requires knowledge of the shock angle β. For example... Figure 2 As shown in (c), the initial value of β4 is first estimated using β1, and then the Newton-Raphson iterative method is used to solve iteratively until convergence. Combining the incoming flow conditions and the flow field parameters at point 1, the shock wave angle at point 1 can be calculated, and its expression is:

[0082]

[0083] In the formula, Ma0 and p0 represent the incoming Mach number and incoming static pressure, respectively. After obtaining β1, the initial value of β4 is estimated to be 1.01 times β1. Once β4 is obtained, point 4 can be solved using the shock wave relation. The calculation formula is:

[0084]

[0085]

[0086]

[0087]

[0088] In the formula, ρ0 represents the incoming flow density, and the subscript 4s represents the parameter at point 4 obtained from the shock wave relation, used to distinguish it from the parameter at subscript 4c obtained through the compatibility equation. The appropriate shock wave angle β4 is solved using the Newton-Raphson iteration method, such that |p 4s -p 4c | To obtain the minimum value. The objective function of this iterative method is: Introduce small perturbations. To aid in the solution, the perturbed objective function is calculated using β₄ + Δβ₄, denoted as N(β + Δβ₄). A difference scheme is used to replace the differential operation of the objective function, expressed as: dN = [N(β₄ + Δβ₄) - N(β₄)] / Δβ 4， The shock angle update formula is: β 4N =β4-N(β4) / dN. Substitute the updated shock angle into the shock relation to calculate the physical parameters at point 4, which are then used as the results of the current iteration step. The slopes of the lines 1-4 and 2-4 are respectively Γ. s =tan2β1+β 4N Given Γ2 = tan2θ4 + μ4 + θ2 + μ2, solve the equations of lines 1-4 and 2-4 simultaneously. Determine the position of point 4 by the intersection point. After recording the residuals and updating the shock angle at point 4, the next iteration cycle can begin.

[0089] The overall shape and spatial location of the curved shock wave can be obtained using the method of rotational characteristic lines. The specific calculation process is as follows: Figure 3 As shown. Figure 3 (a) Curve OP 15 The oblique shock wave generated by the initial external compression surface. Given the inclination angle θ0 of the compression surface, the oblique shock wave angle β0 can be solved using the following formula.

[0090] The flow field parameters behind the shock wave, P, can be solved using the oblique shock wave relation. 11 Point parameters are also included. The compression surface is tangent to the surface at this point, and a C-shaped line is drawn from this point. + The characteristic line intersects the oblique shock wave at P. 15 Points. To keep the illustration simple, only 5 points are marked on the feature line; in the actual solution process, a total of 150 calculation points are set. For example... Figure 3 As shown in (b), recursive calculations are performed starting from these points. The method of characteristics is used to solve for P. 21 P 25The flow field parameters at all points along the line are solved recursively for all downstream points in the same way. The recursive calculation stops when the calculated shock point x-coordinate exceeds 1m. This is because the inlet lip is located at coordinates (1m, 0.4m), and the outer compression section of the inlet is screened according to the shock wave contact criterion, eliminating subcritical and supercritical conditions. The x-coordinate of the terminal shock point is approximately 1m, and 150 calculation nodes are arranged on the initial feature line. The x-coordinate of the terminal shock point is always near 1m. Therefore, it is only necessary to compare the y-coordinate of the terminal shock point with the lip height to determine whether the standard critical state has been reached. The screened samples will be used for neural network training. To enable the neural network to more accurately identify critical state samples, the dataset needs to completely cover all shock wave contact condition samples. Therefore, the y-coordinate of the terminal shock point is set to satisfy the following condition: end Samples ∈ [0.395, 0.402] are defined as qualified samples, which can guarantee that y end The critical sample set with a value of 0.4 is a proper subset of the overall dataset. Samples within this interval are defined as qualified samples under critical operating conditions. Invalid configurations corresponding to subcritical and supercritical conditions are removed, and the effective sample set is finally constructed.

[0091] The number of valid samples obtained varies under different incoming Mach numbers: the dataset sample size and training parameters corresponding to different incoming Mach numbers are shown in Table 2.

[0092]

[0093] According to statistics, the total number of valid samples is less than 10% of the original 1,000 samples, and the sample reduction effect is significant.

[0094] The contour lines of the flow field in the outer compression section obtained by the method of rotational characteristic lines are as follows: Figure 4 As shown, the airflow is smoothly pressurized along the curved surface and the Mach number gradually decreases, demonstrating that this external compression configuration has good compression characteristics.

[0095] To verify the accuracy of the method of characteristics, the solution results were compared with those from viscous numerical simulations. The pressure distribution on the outer compression section wall under different geometric configurations and flow conditions is shown below. Figure 5 As shown in the figure, the wall pressure distribution obtained by the two calculation methods is basically consistent, with small calculation deviations.

[0096] The bending shock wave morphology obtained by the method of characteristics is compared with the CFD numerical schlieren visualization results, for example... Figure 6 As shown, the shock wave profiles of the two samples highly overlap, proving that the spiral characteristic line method can accurately characterize the shock wave morphology and location. The absolute errors between the abscissa of the shock wave tip and the abscissa of the lip are statistically analyzed for all samples, and the results are as follows: Figure 7As shown, the maximum deviation under all operating conditions does not exceed 0.0065m, indicating high reliability of the shock wave position calculation results. It can be guaranteed that the abscissa of the end shock wave point is basically close to the abscissa of the lip. Therefore, by simply comparing the relative positions of their ordinates, it can be determined whether the critical condition for shock wave contact is met.

[0097] The typical critical operating condition samples obtained after screening and the corresponding flow coefficient analysis results are as follows: Figure 8 As shown, whether the intake is in a critical state (whether it meets the "shock wave contact" condition) directly affects the flow capture capability, that is, the flow coefficient φ. Figure 8 Four operating conditions satisfying the method of characteristics screening criteria are given, with the ordinate y of the terminal shock point being... end All fall within the interval [0.395, 0.402]. Among the selected operating conditions, the operating conditions corresponding to the incoming Mach numbers Ma0=4.5 and Ma0=6, y end All are less than 0.4, and all flow coefficients are greater than 0.99; the y values ​​for the other two operating conditions are... end Within the range of [0.4, 0.402], the flow coefficient φ is between 0.97 and 0.98. As can be seen from the schlieren results, the operating conditions near the lower boundary of the screening criterion (y... end The shock wave (y = 0.3954) did not impact the lip wall. Flow coefficient calculations indicate that the operating conditions near the upper boundary of the screening criterion (y = 0.3954) are... end The flow coefficient of samples with a coefficient of 0.4019 is not lower than 0.975. The flow coefficients of samples within the judgment interval all remain at a high level, further verifying that the sample selection criteria adopted in this embodiment are reasonable and feasible.

[0098] Viscous numerical simulations were performed on each intake configuration in the valid sample set to extract the corresponding total pressure recovery coefficient and static pressure rise coefficient, and data annotation was completed for all samples. The annotated valid sample set was then divided into a training set and a validation set in an 8:2 ratio to form the model training dataset.

[0099] The loss function change curve during model training is as follows: Figure 9 As shown, the loss values ​​of the validation set for each operating condition eventually converge to The network training status is stable and the convergence effect is good.

[0100] The relative errors of all samples in the validation set under four operating conditions are summarized in the form of statistical distribution plots. Figure 10 The graph shows the relative error on the horizontal axis and the frequency distribution on the vertical axis. In the pressure rise coefficient prediction results, the relative error of 60% of the samples is less than 1%; for the total pressure recovery coefficient, the prediction error of all samples in the validation set does not exceed 1.6%, and the error of more than half of the samples is less than 0.2%. Figure 10Subfigures (a) and (b) show the comparison between the actual and predicted values ​​of the validation set samples under four incoming flow conditions, along with the corresponding absolute errors. The absolute error of the predicted pressure rise coefficient π is controlled within ±2, while the absolute error of the predicted total pressure recovery coefficient σ is limited to ±0.015. The overall prediction accuracy of the model meets the requirements for engineering use.

[0101] Adjust lip mask control parameters The performance parameters of the air intake were calculated using both numerical simulation and surrogate modeling methods. The results are compared with those of other methods. Figure 11 As shown. Each incoming Mach number Ma0 is matched with fixed compression surface parameters, for example, as... Figure 8 As shown, Ma0=4.5 corresponds to λ1=0.6528. Under all operating conditions, the relative errors between the surrogate model predictions and numerical simulation results are extremely small. Furthermore, the surrogate model accurately characterizes the performance parameter changes caused by adjustments to the lip mask parameters. Under Ma0=4.5 and Ma0=5.0 conditions, as λ2 increases, the pressure rise coefficient π decreases, while the total pressure recovery coefficient σ increases. However, under Ma0=5.5 and Ma0=6 conditions, when λ2>0.2, the pressure rise coefficient π remains relatively stable, while the total pressure recovery coefficient σ continues to increase with λ2. This trend can be reflected in both numerical simulation results and the surrogate model's accurate and effective predictions. It can be seen that the surrogate model's predictions and numerical simulation results show completely consistent trends, accurately tracking performance fluctuations caused by parameter changes.

[0102] Combine different Density gradient contour map of the internal compression section under operating conditions, such as Figure 12 As shown, Figure 12 Two typical operating conditions are presented: Ma0=4.5, θ0=9.2411°, λ1=0.6528 and Ma0=6, θ0=10.7955°, λ1=0.8314. It can be seen that the flow field structure in both conditions changes significantly with the parameter λ2. As λ2 increases, the intensity of the lip shock wave weakens, and the separation bubble (marked by the purple circle) gradually dissipates. For the Ma0=4.5 condition, the oblique shock wave within the shock wave system continuously decays while gradually generating a weak compression wave; while in the Ma0=6 condition, a series of weak compression wave systems exist initially (marked by yellow arrows), and these weak compression waves eventually converge to form a single shock wave (marked by the red arrow). To quantitatively characterize the correlation between the geometric parameter λ2 and performance parameters under this inlet configuration, this method applies a surrogate model to the forward design analysis. Among the three input parameters, θ0 and λ1 are used... Figure 8 The relationship between the set values ​​λ2 and the pressure rise coefficient π and the total pressure recovery coefficient σ is as follows: Figure 13 As shown, Figure 13(a) Ma0=4.5, θ0=9.2411°, λ1=0.6528; Figure 13 (b) Ma0=5.0, θ0=11.0901°, λ1=0.6694; Figure 13 (c) Ma0=5.5, θ0=8.6374°, λ1=0.8658; Figure 13 (d) Ma0=6.0, θ0=10.7955°, λ1=0.8314, where Figure 13 (c) and (d) correspond to high Mach number operating conditions. As λ² increases (the curvature of the curved lip increases), the pressure rise coefficient π increases within a certain parameter range. To investigate this phenomenon, three sets of samples were selected for each operating condition within the corresponding range for numerical simulation. Figure 13 (c) Working condition, λ2 = 0.25 and 0.5 are selected; for Figure 13 (d) For the operating conditions, λ² = 0.34, 0.375, and 0.5 were selected. The numerical simulation results (denoted as CFD in the table) and the surrogate model prediction results (denoted as NN in the table) are listed in Table 3.

[0103]

[0104] Under all operating conditions, the relative prediction error of the neural network surrogate model did not exceed 0.013. Numerical simulation results confirmed that within this range of λ² values, the pressure rise coefficient π does indeed increase with increasing λ². This indicates that the surrogate model can accurately characterize the complete aerodynamic characteristics of the inlet. To explore the above mechanism, Figure 14 The pressure distribution patterns on the inner wall of the lip shield for each sample group are presented. The trends shown by the two incoming flow conditions are similar. As can be seen from the pressure distribution contour plots, increasing λ2 increases the curvature of the lip shield; increased curvature weakens the lip shield shock wave intensity, which can be visually demonstrated from the pressurization characteristics of the lip shield shock wave in the contour plots. Therefore, in the compression and pressurization zone behind the shock wave, the wall static pressure exhibits the following pattern: the static pressure at the same abscissa position is always negatively correlated with the parameter λ2. However, this pattern changes after the airflow enters the expansion and depressurization zone. In the expansion zone downstream of the shock wave, the airflow static pressure gradually decreases; configurations with a larger λ2 value result in a more gradual rate of static pressure decrease, maintaining a higher static pressure level at the outlet. This also explains, mechanistically, why the pressure rise coefficient π shows a positive correlation with λ2 within this parameter range. Figure 14 The static pressure curve in the expansion zone shows that the static pressure of the airflow experienced multiple stages of pressure increase during the decrease (marked by blue arrows). Combined with... Figure 15The density gradient cloud map of the inner compression section can explain this mechanism: weak compression wave reflection and intersection occur in the compression zone behind the shock wave; while in the expansion zone, wave reflection disappears, and the flow field is dominated by compression waves. As the lip curvature increases, the intensity and number of compression waves gradually increase, a characteristic that is more pronounced under Ma0=6 conditions. These compression waves exert a certain pressurizing effect on the airflow in the expansion zone, thereby slowing down the decrease in static pressure, which also reasonably explains... Figure 14 The analysis of the flow field mechanism further validated the rationality of the positive design curve obtained based on the neural network surrogate model.

[0105] The results show that a curved lip induces weak compression waves, which alter the flow field structure and airflow pressurization patterns. The inlet performance parameters do not exhibit a monotonic change trend with increasing lip curvature. The surrogate model trained in this embodiment can fully capture the complex nonlinear correlation between geometric parameters and aerodynamic performance, demonstrating excellent model generalization ability.

[0106] Based on the above verification results, it can be seen that the fully connected neural network proxy model built in this embodiment can realize the rapid mapping from the geometric parameters of the air intake to the aerodynamic performance parameters, and can replace the traditional viscous numerical simulation to complete the rapid prediction of the performance parameters of the hypersonic curved surface air intake.

[0107] Next, refer to the appendix. Figure 16 This paper describes an intake performance parameter prediction model combining the method of characteristics, proposed according to an embodiment of the present invention.

[0108] Figure 16 This is a block diagram of an intake performance parameter prediction model combining the method of characteristics according to an embodiment of the present invention.

[0109] like Figure 16 As shown, the inlet performance parameter prediction model 100 combining the method of characteristics includes:

[0110] The parameterization module 101 is used to perform geometric parameterization on the curved surface compression intake and determine several geometric design parameters that characterize the shape of the intake.

[0111] The initial sample set construction module 102 samples within the range of values ​​of each geometric design parameter to generate multiple sets of air intake geometric configurations, thus forming the initial sample set;

[0112] The effective sample set acquisition module 103 is used to perform inviscid flow field calculations on each inlet configuration in the initial sample set using the method of characteristics, obtain the corresponding shock wave position, and screen the initial sample set according to the shock wave lip judgment criterion to obtain an effective sample set that meets the critical operating conditions.

[0113] The sample labeling module 104 is used to perform viscous numerical simulation on each inlet configuration in the effective sample set, extract the corresponding aerodynamic performance parameters, and complete the sample labeling.

[0114] The partitioning module 105 divides the labeled valid sample set into a training set and a validation set to form a training dataset.

[0115] The training and verification module 106 uses the training dataset to train the neural network and verifies the model accuracy through the verification set to obtain the intake duct performance prediction proxy model.

[0116] It should be noted that the foregoing explanation of an embodiment of an inlet performance parameter prediction method combined with the method of characteristics also applies to the inlet performance parameter prediction model combined with the method of characteristics in this embodiment, and will not be repeated here.

[0117] Figure 17 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. The electronic device may include:

[0118] The memory 201, the processor 202, and the computer program stored on the memory 201 and capable of running on the processor 202.

[0119] When the processor 202 executes the program, it implements the inlet performance parameter prediction method combined with the characteristic line method provided in the above embodiments.

[0120] Furthermore, electronic devices also include:

[0121] Communication interface 203 is used for communication between memory 201 and processor 202.

[0122] The memory 201 is used to store computer programs that can run on the processor 202.

[0123] The memory 201 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage.

[0124] If the memory 201, processor 202, and communication interface 203 are implemented independently, then the communication interface 203, memory 201, and processor 202 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 17 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0125] Optionally, in a specific implementation, if the memory 201, processor 202, and communication interface 203 are integrated on a single chip, then the memory 201, processor 202, and communication interface 203 can communicate with each other through an internal interface.

[0126] The processor 202 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.

[0127] This invention also provides a computer program product, which, when executed by a processor, implements the above-described method for predicting inlet performance parameters using the characteristic line method.

[0128] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for predicting inlet performance parameters using the method of characteristic lines.

[0129] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0130] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0131] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0132] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0133] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0134] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0135] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0136] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for predicting inlet performance parameters using the method of characteristics, characterized in that, Includes the following steps: The curved surface compression inlet is geometrically parameterized to determine several geometric design parameters that characterize the shape of the inlet. Sampling is performed within the range of values ​​for each geometric design parameter to generate multiple sets of intake duct geometric configurations, forming an initial sample set; The inviscid flow field of each inlet configuration in the initial sample set was calculated using the method of characteristics to obtain the corresponding shock wave position. The initial sample set was then screened according to the shock wave lip judgment criterion to obtain an effective sample set that satisfies the critical operating condition. Viscous numerical simulations were performed on each inlet configuration in the effective sample set to extract the corresponding aerodynamic performance parameters and complete the sample labeling. The labeled valid sample set is divided into a training set and a validation set to form a training dataset. The neural network is trained using the training dataset, and the model accuracy is verified using the validation set to obtain the inlet performance prediction surrogate model.

2. The method according to claim 1, characterized in that: The geometric design parameters include a first type of parameters that control the shape of the outer compression surface of the intake duct, and a second type of parameters that control the shape of the intake duct lip cover.

3. The method according to claim 1, characterized in that: The intake duct geometry is generated by using Latin hypercube sampling to generate multiple intake duct geometry configurations.

4. The method according to claim 1, characterized in that: The shock wave position is obtained by solving the supersonic inviscid flow control equation using the swirling characteristic line method, and calculating the position of the curved shock wave corresponding to the inlet.

5. The method according to claim 4, characterized in that, The shock wave lip determination criterion is as follows: when the ordinate of the end of the curved shock wave falls into a preset value range, the current intake configuration is determined to be a valid sample of the critical operating condition.

6. The method according to claim 1, characterized in that, The viscous numerical simulation method is as follows: viscous numerical simulation is performed by solving the Reynolds-averaged Navier-Stokes equations and combining them with a turbulence model; the aerodynamic performance parameters include the total pressure recovery coefficient and the static pressure rise coefficient.

7. The method according to claim 1, characterized in that: The effective sample set is divided into a training set and a validation set in an 8:2 ratio; the neural network adopts a fully connected deep neural network and uses the Adam optimizer and mean squared error loss function to complete the network training.

8. An inlet performance parameter prediction model combining the method of characteristics, characterized in that, include: The parameterization module is used to perform geometric parameterization on the curved surface compression intake, and to determine several geometric design parameters that characterize the shape of the intake. The initial sample set construction module samples within the range of values ​​for each geometric design parameter to generate multiple sets of intake duct geometric configurations, thus forming the initial sample set. The effective sample set acquisition module is used to perform inviscid flow field calculations on each inlet configuration in the initial sample set using the method of characteristics, obtain the corresponding shock wave position, and screen the initial sample set according to the shock wave lip judgment criterion to obtain an effective sample set that meets the critical operating conditions. The sample labeling module is used to perform viscous numerical simulation on each inlet configuration in the effective sample set, extract the corresponding aerodynamic performance parameters, and complete the sample labeling. The partitioning module divides the labeled valid sample set into a training set and a validation set to form the training dataset. The training and verification module trains a neural network using the training dataset and verifies the model accuracy using a verification set to obtain an intake duct performance prediction proxy model.

9. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the method as claimed in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It stores a computer program thereon, the computer program being used to cause the computer to perform the method as described in any one of claims 1-7.

Images