Intake passage compression profile optimization method based on curvature similarity and genetic algorithm

Abstract

The application discloses an air inlet compression profile optimization method based on curvature similarity and a genetic algorithm, and relates to the technical field of aircraft air inlet design. The application fuses CST parameterization, curvature weighted RBF grid deformation and an adaptive genetic algorithm. Through CST fitting, complex profiles are reduced to a small number of coefficients, solving the problem of many control parameters and large design space in traditional parameterization. Through displacement vector and curvature weighted improved thin plate spline kernel function, only control point displacement is used to drive grid smooth deformation, avoiding the problems of large grid reconstruction overhead and inconsistent quality, and deformation in key areas such as the lip is more accurate. Multiple geometric constraints are set to build a penalty function, which is optimized together with the pressure coefficient, to ensure that the profile has excellent aerodynamic performance and meets engineering connection. Based on curvature similarity, the cross probability is dynamically adjusted, genetic algorithm and CFD iteration optimization are combined, the problem that the gradient algorithm is easy to fall into local optimum is overcome, and the global optimization and convergence stability are improved.

Description

Technical Field

[0001] This invention relates to the field of aircraft air intake design technology, and in particular to an air intake compression profile optimization method based on curvature similarity and genetic algorithm. Background Technology

[0002] In the field of aircraft inlet design, aerodynamic optimization of compression profiles has always been a critical and challenging task. Traditional design methods typically rely on empirical formulas, parameter sweeps, or gradient-based local optimization, which often prove inadequate when dealing with wide-range operating conditions. Specifically, existing technologies face several inherent shortcomings: First, in terms of parametric characterization, while traditional methods such as B-splines or NURBS can describe complex profiles, they require numerous control parameters and are difficult to directly embed aerodynamic constraints, resulting in a large design space and low optimization efficiency.

[0003] Furthermore, these parametric methods are not adept at ensuring the smoothness and aerodynamic rationality of the profile during deformation, potentially leading to non-physical interpretations that require repeated intervention and adjustments by designers, significantly increasing labor costs and time. Secondly, in computational fluid dynamics analysis, whenever the profile changes, the traditional approach is to regenerate the computational mesh. This process not only consumes significant computational resources but also struggles to maintain consistent mesh quality. The introduced numerical errors can interfere with the reliability of the optimization process, especially for flow problems like inlet ducts that are highly sensitive to boundary layer meshes; mesh reconstruction often becomes a bottleneck restricting optimization efficiency.

[0004] Finally, at the optimization algorithm level, the performance indicators of the air intake, such as pressure pulsation and total pressure recovery coefficient, typically exhibit strong nonlinearity and multiple extrema characteristics over a wide operating range. Traditional gradient-based optimization algorithms are prone to getting trapped in local optima, while surrogate model-based optimization methods, although reducing computational costs, heavily rely on the quantity and quality of samples for accuracy, facing the "curse of dimensionality" in high-dimensional design spaces. Currently, there is a lack of an efficient optimization strategy that can achieve global optimization at a reasonable computational cost while ensuring solution quality and engineering practicality. Summary of the Invention

[0005] The purpose of this invention is to provide an intake duct compression profile optimization method based on curvature similarity and genetic algorithm, so as to improve the technical problem that the existing technology cannot achieve global optimization at a reasonable computational cost and ensure the quality of optimization results.

[0006] To achieve the above-mentioned objectives, the embodiments of the present invention provide the following technical solutions:

[0007] An inlet compression profile optimization method based on curvature similarity and genetic algorithm includes:

[0008] The original compression profile and compression profile coordinate data of the original intake configuration are collected, and the CST function is used for fitting to generate the compression profile curve and its initial CST coefficients.

[0009] By comparing the original compression profile with the preset initial target compression profile and calculating the displacement vectors of the boundary grid points, the displacement field and deformation grid data of the corresponding domain grid are calculated using the improved thin plate spline kernel function.

[0010] Set the endpoint tangency constraint, convex function constraint, and curvature change constraint of the surface, and construct the penalty function; based on the penalty function and the pressure coefficient, construct the optimization objective;

[0011] Based on the optimization objective and displacement field, the curvature similarity of the original compression surface is calculated, and the initial CST coefficients are iteratively optimized using a genetic algorithm.

[0012] In the above process, this method uses the CST function to fit the original profile, reducing the dimensionality of the complex geometry to a small number of coefficients. This effectively solves the problems of numerous control parameters and large design space in traditional B-spline or NURBS parametric methods, thus improving optimization efficiency. Secondly, by comparing the displacement vectors of the original and target profiles and combining them with an improved thin-plate spline kernel function incorporating curvature weighting, it achieves smooth deformation of the entire computational mesh driven solely by the displacement of boundary grid points. This effectively solves the problems of high computational cost and inconsistent mesh quality caused by traditional mesh reconstruction. In particular, the curvature weighting design makes the deformation of key areas such as the lip more accurate, ensuring the density and rationality of the deformed mesh at geometric features.

[0013] Furthermore, by setting multiple geometric constraints such as endpoint tangency, convexity, and curvature variation to construct a penalty function, which, together with the pressure coefficient, constitutes the optimization objective, this ensures that the optimized profile not only has excellent aerodynamic performance but also meets the engineering requirements of smooth connection with upstream and downstream structures and non-separation flow, avoiding the non-physical aspects that may arise in traditional optimization. Finally, by dynamically adjusting the crossover probability based on curvature similarity and combining iterative optimization with a genetic algorithm, the drawback of gradient algorithms being prone to getting trapped in local optima is effectively overcome. Simultaneously, a reasonable curvature distribution of the profile is preserved, significantly improving global optimization capability and convergence stability. This enables the rapid acquisition of optimized profiles with significantly improved aerodynamic performance under wide operating conditions.

[0014] Furthermore, the calculation of the corresponding displacement field and deformation mesh data includes:

[0015] Set an initial target compression profile and compare it with the original compression profile. Calculate the displacement vector between each boundary grid point in the original compression profile and the corresponding boundary grid point in the initial target compression profile.

[0016] Calculate the curvature of each boundary grid point in the original compressed surface; based on each curvature, construct an improved thin plate spline kernel function;

[0017] Based on each displacement vector and the improved thin plate spline kernel function, the displacement-space relationship and interpolation conditions between boundary grid points are established, and the displacement field between the original compression surface and the initial target compression surface is calculated.

[0018] Based on the displacement field, the original compressed surface mesh data is attenuated by an attenuation function to generate deformed mesh data.

[0019] Furthermore, the formula corresponding to the improved thin-plate spline kernel function is:

[0020] ;

[0021] in, This represents the vertical distance between the compression profile curve and the original compression profile at a certain point on the horizontal axis. Indicates the first Each boundary grid point is at a vertical distance Improved thin-plate spline kernel function values, Indicates the curvature weighting coefficient. Indicates the first Curvature of boundary grid points Absolute value This represents a logarithmic function with a base of a constant.

[0022] In the above process, this invention introduces an improved thin-plate spline kernel function based on the curvature weighting of boundary grid points, and combines it with displacement vectors and attenuation functions to construct a high-precision, adaptive mesh deformation mechanism. Specifically, firstly, by calculating the displacement vector between the boundary grid points of the original and target surfaces, the key driving information of surface changes is accurately captured; then, the kernel function is weighted using the local curvature of each boundary grid point, so that in geometrically critical regions with drastic curvature changes (such as the lip), the displacement influence can be more significantly transmitted to the surrounding grid, thereby ensuring that the deformed mesh maintains density and rationality at geometric features; based on this, by establishing displacement-space relationships and interpolation conditions, accurate interpolation from boundary displacement to the displacement field of the entire computational domain is achieved; finally, a distance-based attenuation function is introduced to effectively suppress non-physical distortions that may occur in distant grids far from the surface, ensuring a smooth transition of the deformation field throughout the entire domain. This series of operations together achieves the effect of driving the overall mesh to undergo automated, high-fidelity deformation solely through boundary point displacement. This completely avoids the problems of high computational overhead and inconsistent mesh quality caused by repeated mesh reconstruction due to surface modifications in traditional methods. At the same time, it significantly improves the adaptability and robustness of mesh deformation at key geometric features, laying a solid foundation for subsequent efficient and reliable iterative optimization.

[0023] Furthermore, the optimization objectives include:

[0024] Based on the throat section tangent angle and inlet section tangent angle of the original compression profile, construct the profile endpoint tangency constraint;

[0025] Based on the compression profile curve, the second derivative value of each sampling point is calculated by the uniform sampling method to construct convex function constraints;

[0026] Calculate the curvature at each sampling point and construct curvature variation constraints;

[0027] Based on the tangency constraint of the surface endpoints, the convex function constraint, and the curvature change constraint, a penalty function is constructed; based on the penalty function and the pressure coefficient, an optimization objective is constructed.

[0028] Furthermore, the formula corresponding to the tangent constraint at the endpoints of the profile is:

[0029] ;

[0030] in, This represents the CST vector composed of the current CST coefficients. The tangency constraint penalty term at the endpoints of the surface is below. , These represent the weighting coefficients for the inlet section tangency penalty and the throat tangency penalty, respectively. , These represent the compression profiles calculated from the current CST coefficients at the starting point. ,end The first derivative at that point, Represents the tangent function. , These represent the tangent angles of the inlet section and the throat section, respectively.

[0031] Furthermore, the formula corresponding to the convex function constraint is:

[0032] ;

[0033] in, Represents the maximum value function. Represents CST vector The convex function constraint penalty term under the following conditions , These represent the x and y coordinates of the sampling points, respectively. Indicates the first The x-coordinate of each sampling point Indicates the compression profile curve at the th The second derivative value at each sampling point Indicates the number of sampling points.

[0034] In the above process, firstly, this method solves the key technical problems of high computational cost and inconsistent mesh quality caused by repeated mesh reconstruction due to changes in the air intake profile in traditional intake optimization. Firstly, by calculating the displacement vector between the boundary mesh points in the original profile and the target boundary mesh points, the deformation boundary conditions are established. Then, an innovative curvature weight coefficient is introduced to construct an improved thin-plate spline kernel function, significantly amplifying the deformation influence in key areas with large curvature, such as the lip, ensuring the density and rationality of the deformed mesh at geometric features. Based on this, by establishing displacement-space relationships and interpolation conditions, the displacement field of the entire computational domain can be accurately interpolated using only the boundary point displacements. Finally, a distance-based attenuation function is introduced to effectively suppress non-physical distortions of the far-end mesh, ensuring a smooth transition of the deformed field.

[0035] Secondly, CST parameterization efficiently reduces the dimensionality of complex geometries, solving the problem of large design space in traditional methods; curvature-weighted RBF mesh deformation completely eliminates the mesh reconstruction bottleneck, ensuring precise synchronization between geometric changes and mesh updates in each iteration; the synergistic effect of the two enables the entire optimization process to not only automatically generate reasonable surfaces that meet engineering constraints such as endpoint tangency and convexity, but also significantly reduce the shoulder pressure coefficient under wide-range working conditions.

[0036] Furthermore, the iterative optimization of each initial CST coefficient using a genetic algorithm includes:

[0037] Based on each initial CST coefficient, an initial population is randomly generated and used as the population in the current iteration;

[0038] Based on the optimization objective, calculate the fitness of each individual in the population in the current iteration; based on the fitness, select the parent individual in the current iteration.

[0039] Calculate the curvature similarity and crossover probability between each pair of parent individuals in the current iteration;

[0040] Based on random numbers and the crossover probability in the current iteration, crossover operations are performed on the population in the current iteration through arithmetic crossover to generate updated offspring individuals in the current iteration.

[0041] Based on each updated offspring individual and displacement field, an updated CAS file is generated and the corresponding pressure coefficient is calculated;

[0042] Based on each updated offspring individual and the stress coefficient, determine whether the convergence condition is met; if so, take the updated offspring individual with the highest fitness as the optimization result; otherwise, iterate again.

[0043] Furthermore, the formula corresponding to the crossover probability is:

[0044] ;

[0045] in, Indicates the crossover probability. , These represent the minimum and maximum crossover probabilities, respectively. Indicates half-value similarity. Indicates curvature similarity.

[0046] Furthermore, the convergence conditions include geometric convergence conditions and pressure convergence conditions;

[0047] The geometric convergence condition is: if the average change of a set of CST coefficients for all individuals in the population under the current iteration compared to the set of CST coefficients under the previous iteration is less than 0.5%, then the geometric convergence condition is satisfied.

[0048] The pressure convergence condition is: if the relative improvement rate of the pressure coefficient at the intake shoulder monitoring point under the current iteration is less than 0.2% over 8 consecutive generations, then the pressure convergence condition is met.

[0049] In the aforementioned process, this method introduces an adaptive crossover strategy based on curvature similarity. This dynamically adjusts the crossover probability according to the similarity of the curvature distribution of parent individuals, effectively protecting the reasonable local curvature features of the air intake profile and avoiding the destruction of excellent geometric structures caused by traditional random crossover. This enhances the algorithm's local search capability and stability. Simultaneously, it employs dual criteria of geometric convergence and pressure convergence to ensure that the optimization process achieves stable optimality in both profile shape and performance dimensions. This solves the problems of traditional genetic algorithms having a single convergence criterion, being prone to premature convergence or oscillation, and significantly improves the reliability and engineering applicability of the optimization results. Furthermore, the CST coefficient, as a design variable, is used to generate a reliable mesh through RBF deformation for CFD evaluation. The evaluation results are fed back to the genetic algorithm to guide evolution, while curvature similarity protects key geometric features. This collaborative mechanism systematically solves the comprehensive problems of low parameterization efficiency, high mesh update cost, poor optimization convergence, and easy destruction of aerodynamic shape in inlet compression profile optimization, achieving a dual breakthrough in aerodynamic performance and design efficiency under wide operating conditions. Attached Figure Description

[0050] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1This is a flowchart of the method in an embodiment of the present invention;

[0052] Figure 2 This is a comparison diagram of the compression profile before and after optimization in an embodiment of the present invention;

[0053] Figure 3 This is a comparison diagram of the Mach number cloud of the compression profile before and after optimization in an embodiment of the present invention. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0055] Please see Figure 1 This embodiment provides an intake duct compression profile optimization method based on curvature similarity and genetic algorithm. Figure 1 The execution entity of the method shown can be a software and / or hardware device. The execution entity of this application can include, but is not limited to, at least one of the following: user equipment, network equipment, etc. User equipment can include, but is not limited to, computers, smartphones, personal digital assistants (PDAs), and the aforementioned electronic devices. Network equipment can include, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of computers or network servers. Cloud computing is a type of distributed computing, consisting of a super virtual computer composed of a group of loosely coupled computers. This embodiment does not impose any limitations on this.

[0056] An inlet compression profile optimization method based on curvature similarity and genetic algorithm includes:

[0057] S1. Collect the original compression profile and compression profile coordinate data of the original configuration of the intake duct, and use the CST function to fit the data to generate the compression profile curve and its initial CST coefficient.

[0058] Specifically, the original compression profile and its coordinate data of the original intake configuration are acquired using CAD software. Then, based on the compression profile coordinate data, a 7th-order endpoint-constrained CST function is used for fitting to generate the corresponding compression profile curve. Finally, based on a preset objective function, the Levenberg-Marquardt algorithm is used to process the compression profile curve, determining the eight CST coefficients that minimize the preset objective function, and these are used as the initial CST coefficients. .

[0059] The formula for the 7th-order endpoint constraint CST function is as follows:

[0060] ;

[0061] ;

[0062] ;

[0063] ;

[0064] in, Represents the normalized x-coordinate, Represents a coordinate function. Represents a class function (used to define the basic geometric type to which the surface belongs). Represents shape functions (used on top of the basic type defined by class functions to describe the specific undulations and curvature details of the surface). , These represent the ordinates of the starting and ending points of the original compressed surface, respectively. , These represent the first index and the second index, respectively. This represents the summation function. Indicates the first Each CST coefficient, The 7th Bernstein polynomial represents the first... Item value.

[0065] In this embodiment, shape function Bernstein polynomials are used as basis functions for the shape functions. The first and second exponents are used to describe the original compression profile at the leading edge / entrance ( =0) bluntness and at the posterior margin / throat ( In this embodiment, the closed-loop property of =1) is utilized. , The values ​​are 0.5 and 1.0, respectively.

[0066] The formula corresponding to the objective function is:

[0067] ;

[0068] ;

[0069] in, Represents the objective function value. , These represent the first and second coordinates of the compressed surface coordinate data, respectively. The x and y coordinates of each boundary grid point Represents the x-coordinate The vertical distance between the compressed profile curve and the original compressed profile. This indicates the number of CST coefficients (i.e.) to Under constraints and with the x-coordinate as The corresponding ordinate. Boundary grid points refer to the grid nodes in the original intake configuration acquired through CAD software, and the coordinates of each grid node are integrated to generate compression surface coordinate data.

[0070] S2. Compare the original compression profile with the preset initial target compression profile and calculate the displacement vector of the boundary grid points. Combine the improved thin plate spline kernel function to calculate the displacement field and deformation grid data of the corresponding domain grid.

[0071] S2 includes:

[0072] S2-1. Set the initial target compression surface and compare it with the original compression surface. Calculate the displacement vector between each boundary grid point in the original compression surface and the corresponding boundary grid point in the initial target compression surface.

[0073] The displacement vector is the position difference between the initial target compression surface and the corresponding boundary grid point of the original compression surface. Taking a boundary grid point in the original compression surface as an example, the difference between the coordinates of the target boundary grid point that matches the boundary grid point in the initial target compression surface and the coordinates of the boundary grid point is used as the displacement vector of the boundary grid point.

[0074] S2-2, Calculate the curvature of each boundary grid point in the original compressed surface; based on each curvature, construct an improved thin plate spline kernel function;

[0075] In this embodiment, since the kernel function value is larger in areas with greater surface curvature (such as the lip), in order to make the deformation of areas with greater curvature (such as the lip) have a more significant impact on the surrounding mesh, thereby ensuring that the deformed mesh is denser at key geometric features, curvature is introduced into the thin plate spline kernel function.

[0076] Taking a boundary grid point as an example, the formula corresponding to the curvature is:

[0077] ;

[0078] in, Indicates the first The x-coordinates of the boundary grid points , They represent the first The first and second derivatives of the x-coordinates of the boundary grid points Indicates the first The curvature of the boundary grid points.

[0079] The formula corresponding to the improved thin-plate spline kernel function is:

[0080] ;

[0081] in, This represents the vertical distance between the compression profile curve and the original compression profile at a certain point on the horizontal axis. Indicates the first Each boundary grid point is at a vertical distance Improved thin-plate spline kernel function values, Indicates the curvature weighting coefficient. Indicates the first Curvature of boundary grid points Absolute value This represents a logarithmic function with a constant base. In this embodiment, the curvature weighting coefficient... The value is 1, thus ensuring that the deformed mesh is denser at key geometric features.

[0082] S2-3. Based on each displacement vector and the improved thin plate spline kernel function, establish the displacement-space relationship and interpolation conditions between boundary grid points, and calculate the displacement field between the original compression surface and the initial target compression surface;

[0083] Specifically, assuming any domain grid point The displacement can be represented by a weighted linear combination of the displacement vectors of all boundary grid points in the original compressed surface. In the x-axis direction, the formula corresponding to the displacement-space relationship between domain grid points and boundary grid points is:

[0084] ;

[0085] in, Represents a grid point in a certain domain displacement in the x direction, Indicates the number of boundary grid points. Indicates the first The weight coefficients corresponding to each boundary grid point Indicates the first The coordinates of the boundary grid points Representation domain grid points With the The Euclidean distance between the x-coordinates of the boundary grid points This represents the improved thin-plate spline kernel function.

[0086] The interpolation conditions are used to solve for the weighting coefficients. The corresponding formula is:

[0087] ;

[0088] in, Indicates the first Displacement vectors of boundary grid points Indicates the first The boundary grid point and the first The Euclidean distance between the boundary grid points.

[0089] The displacement of the domain grid points in the y-axis direction is obtained using the same method as that used to obtain the displacement in the x-axis direction.

[0090] S2-4. Based on the displacement field, the original compressed surface mesh data is attenuated by an attenuation function to generate deformed mesh data.

[0091] If the displacement field is moved directly according to the RBF interpolation, unreasonable distortion may occur in areas far from the profile, or even damage the mesh quality. To ensure that the displacement gradually decreases to zero at the boundary of the affected region and avoids mesh tearing or negative volume caused by step changes, this embodiment uses an attenuation function to attenuate the mesh data of the original compressed profile. This further ensures that the deformation field smoothly transitions to zero at the boundary of the affected region, eliminating step distortion of the mesh. The formula corresponding to the attenuation function in the x-axis direction (and similarly in the y-axis direction) is:

[0092] ;

[0093] in, This represents the attenuated displacement along the x-axis. Indicates the decay index, , These represent the distance and maximum distance from the domain grid point to the surface, respectively.

[0094] The displacement field formed by all attenuated displacements is applied to the mesh data of the original compression surface to generate the corresponding deformed mesh data.

[0095] S3. Set the endpoint tangency constraint, convex function constraint, and curvature change constraint of the surface to construct the penalty function; based on the penalty function and the pressure coefficient, construct the optimization objective;

[0096] S3 includes:

[0097] S3-1. Based on the throat section tangent angle and inlet section tangent angle of the original compression profile, construct the profile endpoint tangency constraint;

[0098] To ensure a smooth connection between the inlet compression profile and adjacent structures (lip section, isolation section) and avoid flow separation, the compression profile must satisfy a given tangential direction constraint at its start and end points. The formula corresponding to the tangential constraint at the profile endpoints is:

[0099] ;

[0100] in, This represents the CST vector composed of the current CST coefficients. The tangency constraint penalty term at the endpoints of the surface is below. , These represent the weighting coefficients for the inlet section tangency penalty and the throat tangency penalty, respectively. , These represent the compression profiles calculated from the current CST coefficients at the starting point. ,end The first derivative (slope) at that point. Represents the tangent function. , These represent the tangent angles of the inlet segment and the throat segment, respectively. In the first iteration, the current CST coefficient is the initial CST coefficient; in the second to nth iterations, the current CST coefficient is the CST coefficient of the previous iteration.

[0101] In this embodiment, the tangent angle of the inlet section The design of the inlet section in the original compression profile obtained from S1 is determined, and =5°; Tangent angle of the laryngeal segment The throat design in the original compression profile obtained from S1 is determined, and =0°.

[0102] S3-2. Based on the compression surface curve, the second derivative value of each sampling point is calculated by the uniform sampling method to construct convex function constraints.

[0103] To ensure the aerodynamic efficiency of the inlet compression profile and prevent flow separation, the compression profile must maintain downward convexity throughout the entire compression section. Since checking an infinite number of points over a continuous interval is impractical, a discrete sampling method is employed, within the interval... Take evenly inside One point (in this embodiment, (Taking the value 21) Checking the convexity condition, the formula corresponding to the convex function constraint is:

[0104] ;

[0105] in, Represents the maximum value function. Represents CST vector The convex function constraint penalty term under the following conditions , These represent the x and y coordinates of the sampling points, respectively. Indicates the first The x-coordinate of each sampling point Indicates the compression profile curve at the th The second derivative value at each sampling point.

[0106] S3-3. Calculate the curvature of each sampling point and construct curvature variation constraints;

[0107] To avoid convexity disruption between sampling points, additional checks are needed on curvature sign changes; therefore, curvature variation constraints are set in this embodiment. The formula corresponding to the curvature variation constraint is:

[0108] ;

[0109] in, express, , They represent the first time. sampling points and the sampling points The curvature of the sampled points is calculated using the same method as that used to calculate the boundary grid points.

[0110] S3-4. Based on the tangency constraint of the surface endpoints, the convex function constraint, and the curvature change constraint, construct the penalty function; based on the penalty function and the pressure coefficient, construct the optimization objective.

[0111] Specifically, according to the formula:

[0112] ;

[0113] in, Represents the penalty function. , These represent the first penalty weight and the second penalty weight, respectively (adjusted according to the importance and sensitivity of the constraint).

[0114] Then, according to the formula:

[0115] ;

[0116] in, This represents the target weight (in this embodiment, the value is 1000). Indicate the optimization objective The minimum value, This represents the pressure coefficient at the shoulder monitoring point under the current CST coefficient. The current CST coefficient follows CST coefficient constraints:

[0117] ;

[0118] Indicates the first The current CST coefficient, Indicates the first Each CST coefficient, This represents the set of CST coefficient constraints generated after transforming all the initial CST coefficients.

[0119] S4. Based on the optimization objective and displacement field, calculate the curvature similarity of the original compression surface, and iteratively optimize each initial CST coefficient using a genetic algorithm (arithmetic crossover).

[0120] S4 includes:

[0121] S4-1. Based on each initial CST coefficient, randomly generate an initial population and use it as the population in the current iteration;

[0122] Specifically, an initial population is generated uniformly and randomly within a range of ±50% (CST coefficient constraint) centered on each initial CST coefficient. Each individual in the initial population has a set of 8 CST coefficients.

[0123] S4-2. Based on the optimization objective, calculate the fitness of each individual in the population under the current iteration; based on the fitness, select the parent individual under the current iteration.

[0124] Specifically, individuals with the top 10% fitness are directly introduced into the next generation without participating in crossover, while the remaining individuals with the bottom 90% fitness are used as parent individuals.

[0125] S4-3. Calculate the curvature similarity and crossover probability between each pair of parent individuals in the current iteration;

[0126] Taking two parent individuals as an example, the formula for curvature similarity is:

[0127] ;

[0128] in, Indicates curvature similarity. , Let represent the maximum and minimum curvature of the population in the current iteration, respectively. , These represent the two parent individuals in the [number]th generation. sampling points The curvature.

[0129] The formula for the crossover probability is:

[0130] ;

[0131] in, Indicates the crossover probability. , These represent the minimum and maximum crossover probabilities, respectively. This represents the half-value similarity (used to control the steepness of the function).

[0132] S4-4. Based on random numbers and the crossover probability in the current iteration, perform crossover operations on the population in the current iteration through arithmetic crossover to generate updated offspring individuals in the current iteration.

[0133] Specifically, random numbers are generated within the interval [0.1]. If the random number is less than the crossover probability in the current iteration, arithmetic crossover is used to crossover individuals with fitness in the bottom 90% of the current iteration; otherwise, the first and second parent individuals are directly used as the updated child individuals.

[0134] S4-5. Based on each updated offspring individual and displacement field, generate an updated CAS file and calculate the corresponding pressure coefficient;

[0135] Specifically, whenever the genetic algorithm gives a new CST coefficient, only after generating a new intake domain mesh through the displacement field and exporting it as a CAS file can the CFD software be called to perform calculations and return the pressure coefficient of the intake shoulder monitoring point. Only in this way can the fitness of the new individual be calculated and new offspring be generated through the genetic algorithm.

[0136] Based on a set of CST coefficients corresponding to each updated offspring individual, the same CST function as in S1 is used for fitting to generate updated compression profile curves. Then, the updated compression profile curves are processed using the same method as in S2 to generate corresponding updated deformable mesh data. A CFD solver is then used to process the updated deformable mesh data to calculate the pressure coefficients at the corresponding inlet shoulder monitoring points.

[0137] S4-6. Based on each updated offspring individual and the stress coefficient, determine whether the convergence condition is met. If so, take the updated offspring individual with the highest fitness as the optimization result; otherwise, increment the iteration count by 1 and return to S4-2, until the convergence condition is met. The initial value of the iteration count is 1.

[0138] Convergence conditions include geometric convergence conditions and pressure convergence conditions, and both geometric convergence conditions and pressure convergence conditions must be satisfied simultaneously.

[0139] The geometric convergence condition is as follows: if the average change of a set of 8 CST coefficients of all individuals in the population under the current iteration is less than 0.5% compared with the set of 8 CST coefficients under the previous iteration, then the geometric design is considered to be stable and the geometric convergence condition is satisfied.

[0140] The pressure convergence condition is as follows: when the pressure coefficient of the intake shoulder monitoring point under the current iteration has a relative improvement rate of less than 0.2% over 8 consecutive generations, the performance is considered to have converged to the optimal level and is considered to meet the pressure convergence condition.

[0141] In summary, this invention utilizes the CST function to accurately fit the original profile, reducing high-dimensional geometric features to a small number of design variables, thus solving the problems of large design space and low optimization efficiency in traditional parametric methods. Subsequently, in each optimization iteration, an RBF interpolation field is constructed by introducing an improved thin-plate spline kernel function with curvature features. This enables the smooth deformation of the entire computational grid to be driven solely by the displacement of boundary grid points, completely avoiding the computational overhead and numerical noise caused by repeated grid reconstruction in traditional methods, and ensuring the consistency of grid quality. Finally, the genetic algorithm aims to minimize the shoulder pressure coefficient and performs global optimization by coupling geometric constraints such as endpoint tangency and curve convexity. At the same time, an adaptive crossover strategy based on curvature similarity is introduced, which effectively protects the reasonable curvature distribution of the profile, avoids random crossover from damaging aerodynamic performance, and reduces the shoulder pressure coefficient of the inlet under corresponding operating conditions.

[0142] A sample intake duct structure is collected to construct the corresponding original compression profile. This original compression profile is then processed using the method described in this embodiment to generate a set of optimized CST coefficients. Based on these optimized CST coefficients, the corresponding optimized compression profile is reconstructed. For example... Figure 2 As shown, the optimized compression profile becomes smoother and gentler in the shoulder area, effectively improving the airflow state; the tangent angles at the start and end points strictly meet the preset constraints, ensuring smooth connection with the upstream and downstream structures; the entire profile maintains good downward convexity, avoiding local depressions or inflection points.

[0143] Simulating the airflow states of the original and optimized compression profiles, the generated Mach number contour plots are as follows: Figure 3 As shown, the separation zone corresponding to the compression profile is significantly reduced after optimization by the method in this embodiment.

[0144] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0145] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for optimizing the compression profile of an air intake based on curvature similarity and a genetic algorithm, characterized in that, include: The original compression profile and compression profile coordinate data of the original intake configuration are collected, and the CST function is used for fitting to generate the compression profile curve and its initial CST coefficients. By comparing the original compression profile with the preset initial target compression profile and calculating the displacement vectors of the boundary grid points, the displacement field and deformation grid data of the corresponding domain grid are calculated using the improved thin plate spline kernel function. Set the endpoint tangency constraint, convex function constraint, and curvature change constraint of the surface, and construct the penalty function; based on the penalty function and the pressure coefficient, construct the optimization objective; Based on the optimization objective and displacement field, the curvature similarity of the original compression surface is calculated, and the initial CST coefficients are iteratively optimized using a genetic algorithm. The calculated displacement field and deformation mesh data include: Set an initial target compression profile and compare it with the original compression profile. Calculate the displacement vector between each boundary grid point in the original compression profile and the corresponding boundary grid point in the initial target compression profile. Calculate the curvature of each boundary grid point in the original compressed surface; based on each curvature, construct an improved thin plate spline kernel function; Based on each displacement vector and the improved thin plate spline kernel function, the displacement-space relationship and interpolation conditions between boundary grid points are established, and the displacement field between the original compression surface and the initial target compression surface is calculated. Based on the displacement field, the original compressed surface mesh data is attenuated by an attenuation function to generate deformed mesh data; The formula corresponding to the improved thin-plate spline kernel function is: ； in, This represents the vertical distance between the compression profile curve and the original compression profile at a certain point on the horizontal axis. Indicates the first Each boundary grid point is at a vertical distance Improved thin-plate spline kernel function values, Indicates the curvature weighting coefficient. Indicates the first Curvature of boundary grid points Absolute value Represents a logarithmic function with base constant; The iterative optimization of each initial CST coefficient using a genetic algorithm includes: Based on each initial CST coefficient, an initial population is randomly generated and used as the population in the current iteration; Based on the optimization objective, calculate the fitness of each individual in the population in the current iteration; based on the fitness, select the parent individual in the current iteration. Calculate the curvature similarity and crossover probability between each pair of parent individuals in the current iteration; Based on random numbers and the crossover probability in the current iteration, crossover operations are performed on the population in the current iteration through arithmetic crossover to generate updated offspring individuals in the current iteration. Based on each updated offspring individual and displacement field, an updated CAS file is generated and the corresponding pressure coefficient is calculated; Based on each updated offspring individual and the stress coefficient, determine whether the convergence condition is met; if so, take the updated offspring individual with the highest fitness as the optimization result; otherwise, iterate again. The formula corresponding to the crossover probability is: ； in, Indicates the crossover probability. , These represent the minimum and maximum crossover probabilities, respectively. Indicates half-value similarity. Indicates curvature similarity.

2. The inlet compression profile optimization method based on curvature similarity and genetic algorithm according to claim 1, characterized in that, The optimization objectives include: Based on the throat section tangent angle and inlet section tangent angle of the original compression profile, construct the profile endpoint tangency constraint; Based on the compression profile curve, the second derivative value of each sampling point is calculated by the uniform sampling method to construct convex function constraints; Calculate the curvature at each sampling point and construct curvature variation constraints; Based on the tangency constraint of the surface endpoints, the convex function constraint, and the curvature change constraint, a penalty function is constructed; based on the penalty function and the pressure coefficient, an optimization objective is constructed.

3. The inlet compression profile optimization method based on curvature similarity and genetic algorithm according to claim 2, characterized in that, The formula corresponding to the tangent constraint at the endpoints of the profile is: ； in, This represents the CST vector composed of the current CST coefficients. The tangency constraint penalty term at the endpoints of the surface is below. , These represent the weighting coefficients for the inlet section tangency penalty and the throat tangency penalty, respectively. , These represent the compression profiles calculated from the current CST coefficients at the starting point. ,end The first derivative at that point, Represents the tangent function. , These represent the tangent angles of the inlet section and the throat section, respectively.

4. The inlet compression profile optimization method based on curvature similarity and genetic algorithm according to claim 2, characterized in that, The formula corresponding to the convex function constraint is: ； in, Represents the maximum value function. Represents CST vector The convex function constraint penalty term under the following conditions , These represent the x and y coordinates of the sampling points, respectively. Indicates the first The x-coordinate of each sampling point Indicates the compression profile curve at the th The second derivative value at each sampling point Indicates the number of sampling points. This represents the summation function.

5. The inlet compression profile optimization method based on curvature similarity and genetic algorithm according to claim 1, characterized in that, The convergence conditions include geometric convergence conditions and pressure convergence conditions; The geometric convergence condition is: if the average change of a set of CST coefficients of all individuals in the population under the current iteration compared with the set of CST coefficients under the previous iteration is less than 0.5%, then the geometric convergence condition is satisfied. The pressure convergence condition is: if the relative improvement rate of the pressure coefficient at the intake shoulder monitoring point under the current iteration is less than 0.2% over 8 consecutive generations, then the pressure convergence condition is met.

Images