A method and system for optimizing lift-drag ratio of airfoils based on design space excitation distribution

Abstract

The application provides a wing profile lift-drag ratio optimization method and system based on design space excitation distribution, which comprises the following steps: parameterizing a wing profile to be optimized, and obtaining a description form of the wing profile optimization problem; selecting sample points in the wing profile design space, calculating the lift-drag ratio of the wing profile described by the sample points, and establishing a global radial basis interpolation model and a local radial basis interpolation model of the design variables to the target variables; using a teach-and-learn optimization method and a social particle swarm optimization method for optimization; using a design space excitation distribution strategy to drive the two optimizers to work in coordination, and generating a preselected population in the design space; using a complex preselection strategy to update the preselected population, supplementing samples after calculation, and updating the model until the optimization is ended after the convergence condition is reached; and determining the wing profile design parameters through exploration and development of the design parameter space that can represent the wing profile shape, thereby saving the number of simulation evaluation times, reducing the calculation cost, and ensuring the optimization quality.

Description

Technical fields:

[0001] This invention belongs to the field of wind airfoil design technology, specifically relating to an airfoil lift-to-drag ratio optimization method and system based on design space excitation allocation, used to perform extreme optimization design of the lift-to-drag ratio performance of a reference airfoil. Background technology:

[0002] The aerodynamic performance of airfoils has a crucial impact on aircraft performance. Designers frequently optimize airfoil geometry to improve aerodynamic parameters, including lift coefficient (L), drag coefficient (D), and lift-to-drag ratio (L / D). In early airfoil design practices, designers primarily modified airfoil geometry based on their own knowledge and experience, which were subjective and varied from designer to designer. Furthermore, as the complexity of airfoil design and performance requirements have increased, the shortcomings and drawbacks of relying on experience and traditional optimization techniques have become increasingly apparent. Therefore, fully exploring the airfoil design parameter space and combining it with CFD calculations can significantly improve airfoil design capabilities and enhance performance indicators. However, the high cost of CFD calculations is unavoidable; therefore, obtaining the optimal airfoil design parameters with minimal computation is crucial.

[0003] Airfoil design parameters often range from tens to hundreds of dimensions, all of which influence the final performance of the airfoil and interact with each other, exhibiting strong nonlinearity. Traditional optimization methods struggle to understand these complex relationships, often only achieving local optima within a design space. This is due to the intricate functional relationships between design and target parameters, making knowledge-based and experience-based design methods inadequate for handling such situations. Metaheuristic intelligent optimization, however, is rapidly developing and is being widely applied to complex nonlinear optimization problems.

[0004] Because CFD calculations and wind tunnel experiments are costly, yet unavoidable in airfoil evaluation, a significant portion of airfoil optimization costs are consumed in airfoil evaluation. Using existing evaluations to progressively predict optimal airfoil design parameters is an intuitive and effective approach; however, limitations in optimization algorithms and surrogate model techniques make cost control and optimization effectiveness difficult to guarantee. Summary of the Invention:

[0005] To address the shortcomings of existing technologies, this invention proposes an airfoil lift-to-drag ratio optimization method based on design space excitation allocation. By exploring and developing the parameter design space that can represent the airfoil shape, the airfoil design parameters are determined, reducing the number of simulation evaluations. This method can effectively solve the shortcomings of current technologies, save costs, and improve optimization efficiency.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: an airfoil lift-to-drag ratio optimization method based on design space excitation allocation, wherein the airfoil aerodynamic characteristic optimization method includes the following steps:

[0007] S1. Parameterize the reference airfoil shape that needs to be optimized, obtain the mathematical description of the design space of the reference airfoil shape, and construct the description of the airfoil optimization problem.

[0008] S2, Select sample points in the design space based on the description form, and perform CFD calculations on the airfoil described by the sample points to obtain the lift-to-drag ratio, and establish a preliminary global radial basis interpolation model and a local radial basis interpolation model from the design variables to the target variables.

[0009] S3 utilizes two optimizers working on the global radial basis interpolation model and the local radial basis interpolation model to generate a preselected population in the design space. The two optimizers are driven by the incentive allocation strategy of the design space.

[0010] S4. Use a complex pre-selection strategy to update the pre-selection population, perform CFD calculations, and then update the global radial basis interpolation model and the local radial basis interpolation model until the results converge, end the optimization, and obtain the optimized airfoil lift-to-drag ratio.

[0011] Furthermore, in S1, for the reference airfoil that needs optimization, the CST parameterization method is used to parameterize the airfoil surface, and d CST parameters are used as variables for airfoil optimization design. Thus, the design variables are determined as: x = [x1, x2, ..., x...]. d ], where xi is the i-th design component of the design variables: i = 1, 2, ..., n; and the formulation of the airfoil lift-to-drag ratio optimization problem is obtained:

[0012]

[0013] subject tot0-t max (x)≤0

[0014]

[0015] i = 1, 2, ..., d

[0016] Where D(x) and L(x) represent the drag coefficient and lift coefficient, respectively, and t0, t max These represent the baseline airfoil thickness and the optimized airfoil thickness, respectively, in the second constraint. This represents the design variable value for the i-th dimension corresponding to the reference airfoil.

[0017] Furthermore, in S2, sample points in the design space are selected based on the mathematical description, and CFD calculations are performed on the airfoil described by the sample points to obtain the lift-to-drag ratio. When obtaining the model from the initial design variables to the target variables, Latin hypercube design is used to extract s samples in the design space: x 1 ,x 2 ,…,x s CFD calculations were performed on the airfoil it represents. Assuming initial conditions, the calculated target variable—the ratio of drag to lift—is: f(x) 1 ),f(x 2 ),…,f(x s The design variables and the corresponding target variables form a sample set, and a global radial basis interpolation model for the sample set is established using radial basis interpolation with a cubic function as the kernel function.

[0018] Furthermore, in S3, the first optimizer is a global radial basis interpolation optimization driven by the teaching and learning optimization algorithm, where the global radial basis interpolation model is constructed from all samples in the sample set; the second optimizer is a local radial basis interpolation optimization driven by the social learning particle swarm optimization algorithm, where the local radial basis interpolation is a local radial basis interpolation model built from the better samples in the sample set, and the range of its model is limited. When one optimizer works on the global radial basis interpolation model or the local radial basis interpolation model, the final pre-selected population is obtained after M iterations, during which the other optimizer is idle.

[0019] Furthermore, in S3, the coordinated work of the two optimizers is driven by the incentive allocation strategy of the design space. The incentive allocation strategy scores the optimizers after each new sampling and uses the optimizer with the higher score in the next round of work.

[0020] Furthermore, in S3, the incentive allocation strategy first designs the contribution of new sampling to exploration and development, and designs the evaluation score of the optimizer based on the contributions of both aspects. After every M iterations, the evaluation score of the working optimizer is updated, but the evaluation score of the idle optimizer is not updated.

[0021] Furthermore, in S4, a dimension-based permutation strategy is implemented for the pre-selected population generated by the optimizer to expand the pre-selected pool. The global radial basis interpolation model and the local radial basis interpolation model of the current optimizer are used to calculate all points in the pre-selected pool. The parameter combination corresponding to the point with the minimum calculated value is selected for airfoil CFD calculation. The calculation conditions are the initial conditions. The design parameters and calculation results are stored in the sample set, and the corresponding evaluation scores are updated according to the calculation results. Then, the process returns to S3 until the calculation results reach the convergence condition.

[0022] Based on the same concept as the method, the present invention provides an airfoil lift-to-drag ratio optimization system based on design space excitation allocation, including an optimization problem construction module, a model construction module, a driving module, and an optimization module; the optimization problem construction module is used to parameterize the reference airfoil shape to be optimized, obtain the mathematical description form of the design space of the reference airfoil shape, and construct the description form of the airfoil optimization problem.

[0023] The model building module selects sample points in the design space based on the description form, and performs CFD calculations on the airfoil described by the sample points to obtain the lift-to-drag ratio, and establishes a preliminary global radial basis interpolation model and a local radial basis interpolation model from the design variables to the target variables.

[0024] The driving module utilizes two optimizers to work on the global radial basis interpolation model and the local radial basis interpolation model to generate a pre-selected population in the design space. The two optimizers are driven by the incentive allocation strategy of the design space.

[0025] The optimization module uses a complex pre-selection strategy to update the pre-selection population, performs CFD calculations, and then updates the global radial basis interpolation model and the local radial basis interpolation model until the results converge, ending the optimization and obtaining the optimized airfoil lift-to-drag ratio.

[0026] Another computer device is provided, including a processor and a memory. The memory is used to store a computer-executable program. The processor reads the computer-executable program from the memory and executes it. When the processor executes the program, it can implement the airfoil lift-to-drag ratio optimization method based on design space excitation allocation described in this invention.

[0027] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can implement the airfoil lift-to-drag ratio optimization method based on design space excitation allocation described in the present invention.

[0028] Compared with existing technologies, the airfoil lift-to-drag ratio optimization method based on design space excitation allocation provided by this invention has the following advantages:

[0029] By exploring and developing the design parameter space that represents the airfoil shape, the airfoil design parameters are determined, saving the number of simulation evaluations and making full use of the airfoil evaluation data each time, thereby saving airfoil design costs. The excitation allocation strategy of the airfoil design space provided by this invention can realize intervention and design throughout the entire airfoil optimization process and can be adjusted according to needs. The lift-to-drag ratio optimization method provided by this invention can be used indiscriminately for the optimization of other airfoil target parameters. The optimized airfoil designed using this invention has extremely strong lift-to-drag ratio performance, approaches the global optimum in the design parameter space, and its optimization level has significant advantages compared with current airfoil optimization. Attached image description:

[0030] Figure 1 This is a flowchart illustrating the specific implementation of the present invention.

[0031] Figure 2 This is a schematic diagram illustrating the design of the spatial excitation framework in this invention.

[0032] Figure 3 It is the lift-to-drag ratio convergence process in the airfoil optimization process.

[0033] Figure 4 These are the optimized airfoil obtained by this invention and a comparison diagram. Detailed implementation method:

[0034] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific examples. These examples are merely illustrative and not intended to limit the invention.

[0035] like Figure 1 As shown, the present invention provides an airfoil lift-to-drag ratio optimization method based on design space excitation allocation, comprising the following steps:

[0036] 1. Airfoil parameterization and initial sampling using Latin hypercube.

[0037] Taking the NACA0012 airfoil as an example, the CST parameterization method is used to parameterize the reference airfoil, obtaining the design variables as x = [x1, x2, ..., x d ], where d is the number of design components. The number of parameters varies for different orders of parametric methods; in this example, d = 14. The range of design variables is set under the parameters of the reference airfoil as follows:

[0038]

[0039] in This represents the design variable value in the i-th dimension corresponding to the reference airfoil. S points x are sampled using the Latin hypercube design in the design parameter space. 1 ,x 2 ,…,x s (In this example, s = 50).

[0040] 2. Perform CFD evaluation and store the data in a database (DB).

[0041] The lift-to-drag ratio was evaluated at the sample points. In this example, CFD calculations were used for the evaluation. Under the design conditions, the free-flow Mach number was 0.50, the airfoil angle of attack was 2°, and the Reynolds number Re = 5.0 × e 6 After initial sampling, such evaluations are performed s times to obtain the target variable values ​​f(x) corresponding to s samples. 1 ),f(x2 ),…,f(x s The data is stored in a database and sorted in descending order of the boost-to-drag ratio to obtain the modeling dataset. In subsequent sampling processes, a unique new sample is evaluated each time and added to the dataset database.

[0042] 3. Perform airfoil optimization based on design space excitation allocation

[0043] This step is the core step of the invention. Driven by the design space incentive allocation framework, the two optimizers work in coordination to complete the preliminary pre-selection of the expected optimal point. It consists of three parts: optimizer 1, optimizer 2, and the design space incentive allocation framework. The specific steps are as follows:

[0044] 3.1 Optimization of Global Radial Basis Interpolation Model Driven by Teaching and Learning Optimization Algorithm (TLBO)

[0045] When stag = 1, a global radial basis function (DB) interpolation model is established using DB. The formula for establishing the global radial basis function interpolation model is as follows:

[0046]

[0047] Where a i For the coefficients that need to be determined, It's a kernel function. p(x) is a linear polynomial that needs to be determined. The above information is determined by solving the following equation:

[0048]

[0049] in, P is a d-dimensional first-order polynomial, and its coefficients form a vector b. F is a column vector consisting of the lift-to-drag ratios of all samples.

[0050] The first s samples in the database are selected as the initial airfoil parameter population for TLBO optimization. During the teaching phase, the airfoil parameter optimization follows the formula:

[0051]

[0052] Where, r i It is a random number between 0 and 1, x T It is the parameter combination with the highest lift-to-drag ratio in the current group, T F It is 1 or 2, randomly selected during the optimization process, where M is the average of the parameters of each component in the current population. If the obtained... Compared to x i If a higher lift-to-drag ratio is achieved, then x will be... i Replace with It should be noted that, except for the lift-to-drag ratio evaluated by CFD for all parameter combinations in the first round of optimization, the lift-to-drag ratios for subsequent updated parameter combinations are all predicted values ​​given by the global radial basis interpolation model. During the learning phase, the optimization of airfoil parameters follows the formula:

[0053] If f(x) i )<f(x j If x i new =x i +r i ×(x i -x j )

[0054] If f(x) i )≥f(x j If x i new =x i +r i ×(x j -x i ).

[0055] The teaching and learning phases alternate, and after a total of several iterations, the final set of design parameter combinations is obtained in preparation for the next step. In this example, there are 30 iterations.

[0056] 3.2 Optimization of Local Radial Basis Interpolation Model Driven by Social Learning Particle Swarm Optimization (SLPSO) Algorithm

[0057] When stag = -1, the first s samples in the database are selected as the initial airfoil parameter population, and SLPSO optimization begins. In each iteration, the d neighboring individuals of each individual in the population are selected from the sample set, with Euclidean distance as the criterion for determining neighboring individuals. This forms the modeling dataset, a local radial basis interpolation model is established, and the search range is determined.

[0058] U=[lb-0.1×(ub-lb),ub+0.1×(ub-lb)]

[0059]

[0060] Then, the social particle swarm optimization algorithm is used for optimization. The population update formula is as follows:

[0061]

[0062] in, t represents the t-th iteration, and j represents x. i In the j-th dimension, r1, r2, r3 are all random numbers uniformly distributed in [0, 1], x k They are randomly selected individuals, and It is the average of all individuals in the population along the j-th dimension. This setting maintains a balance between convergence speed and search accuracy, and the inertia term improves the population diversity of the algorithm. Overall, SLPSO improves search capability and reduces the risk of the regularized particle swarm optimization algorithm getting trapped in local optima.

[0063] Social learning particle swarm optimization (SLPSO) driven airfoil optimization includes the following steps;

[0064] 1) Set the algorithm parameters and initialize the airfoil parameter group;

[0065] 2) When there are fewer than M iterations, rank the boost-to-drag ratio of each parameter combination in descending order;

[0066] 3) Update the population using the formula described above;

[0067] 4) Select d neighboring individuals from the population for each individual in the database to form a modeling dataset and update the local radial basis interpolation model;

[0068] 5) After several iterations, the final set of design parameter combinations is obtained in preparation for the next step. In this example, the iterations are 30.

[0069] 3.3 Design of a spatial incentive allocation framework to drive two optimizers

[0070] The innovation of this invention also lies in providing a design space incentive allocation framework that scores two optimizers and uses the optimizer with the higher score for the next pre-selection point. The design space incentive allocation framework first evaluates the contribution to the exploration and development of the airfoil design space, using this as an incentive for the optimizer. Then, it provides feedback based on the contribution of the current fill sample and uses the overall feedback to form the current optimizer's evaluation score (called Sc). In the next loop, the optimizer with the larger Sc is then executed. In this design space incentive allocation framework, changes in the feedback affect Sc, further influencing the incentive allocation for exploration and development. Several points need to be emphasized: First, for each supplementary sample, the contribution includes both exploration and development contributions; therefore, the feedback should include both aspects. Second, the needs for exploration and development change significantly at different stages of the optimization process. This is because initially, there is no overall grasp of the airfoil design space, so thorough exploration is needed to obtain the overall relationship between design variables and target variables. As optimization progresses, more and more samples accumulate, and the region where the optimal design exists is determined. Therefore, the focus should be on development within this region to efficiently find the parameters of the optimal design. Thus, the feedback on sample contributions in these two aspects should also follow this change. Finally, the score Sc should include past contributions, not just recent ones. For these purposes, this step of the invention first designs the feedback equation. Assume there are N existing samples in DB, and the maximum number of samples is preset to K (K = 300 in this example). The equation determines whether the optimizer generating the i-th supplementary sample will obtain a value named EI due to development. i The feedback depends on the combination of airfoil parameters x found by the optimizer. * Does it satisfy: f(x) * )≤f(x),x∈DB, where f is D(x) / L(x),EI i The calculation method is as follows:

[0071]

[0072] On the other hand, if supplementary samples increase the diversity of the sample set DB, the optimizer will gain feedback EP from exploration. i The criteria for judging diversity have been expanded to include: EP i The calculation method is as follows:

[0073]

[0074] The above conditions require that the supplementary samples should not be too similar to the existing samples, which is to ensure diversity, and the optimizer's score is based on EI. i and EP i Given:

[0075] Sc=0.5×Sc+δ1×EI i +δ2×EP i

[0076] Retain half of the previous Sc to represent past contributions. δ is a function that equals 1 when the condition is met and 0 when the condition is not met. Note that both optimizers have Sc, but the optimizer's Sc is only updated after that optimizer is executed.

[0077] To roughly match the changing exploration and development needs during the search process, a monotonic power function was chosen to represent EP and EI. Extensive testing determined that α should be 2. The formulas for calculating EP and EI show that the sum of EP and EI is 1. A larger EI ratio will encourage the use of optimizers with strong development capabilities; otherwise, it will encourage the use of optimizers with strong development capabilities. During the search process, the incentive allocation for exploration and development will continuously change as N increases. To further illustrate the role of the spatial incentive allocation framework, Figure 2 The process of EP changing with N is shown, denoted as the EP curve, where Ar1 represents the lightly shaded area and Ar2 represents the heavily shaded area. During initial sampling, the aim is to obtain a uniformly distributed sample within the airfoil design space, therefore EP = 1. However, during the filling sampling process, EP gradually decreases, and the exploration motivation diminishes. Due to the boost from EP, space exploration is stimulated when an optimizer is used in the next iteration. Because the optimization search is large and random, this motivation is difficult to quantify, but generally, the magnitude of the EP value is positively correlated with the probability (PE) of a sample being used for exploration.

[0078] EP∝PE

[0079] The expected value of EP is calculated as follows:

[0080]

[0081] Since PE represents the probability of a sampling point being used for exploration, then E(PE) represents the number of sampling points used for exploration, denoted as NP. Therefore, we can obtain: Ar1 ∝ NP. The above derivation also applies to EI. Let NI be the number of sampling points used for development, and then Ar2 ∶ NI. The EP curve, by adjusting the magnitude and distribution of Ar1 and Ar2, largely intervenes in the allocation between exploration and development. Therefore, choosing an appropriate EP curve will help balance exploration and development within the airfoil design space, thereby improving the efficiency of airfoil optimization design.

[0082] 4. Process the group of airfoil parameters generated by the optimizer and produce a parameter combination that requires CFD calculation.

[0083] from Figure 1It can be seen that the design parameter groups generated by the two optimizers both need to undergo multiple screening processes. As another innovation of this invention, a pre-screening strategy is provided to generate the final airfoil design parameter combination x used for CFD calculation. * The pre-selection strategy only applies to the last generation of the population after each restart. First, the top q individuals in the population are selected to form a subpopulation Q. Then, β dimensions are randomly selected. Individuals in Q rotate on each selected dimension, where rotation is an algebraic method that changes the array's order. For example, with q = 5, let τ = (3 1 5 2 4).

[0084] The arrangement of individuals in Q along dimension j can be represented as:

[0085]

[0086] The steps for filtering the group of design parameters generated by the optimizer are as follows:

[0087] 1) Sort individuals in the population in descending order of lift-to-drag ratio (predicted by the RBF model);

[0088] 2) Randomly select an integer q from 1 to s, and select the first q parameters from the population to form Q;

[0089] 3) Randomly select an integer β from 1 to d, and randomly select β from d dimensions;

[0090] 4) Perform rotation operations on each of the above dimensions for individuals in Q and generate new subpopulations. The rotations are also generated randomly.

[0091] 5) Merge the β subgroups with Q to form a new design parameter group;

[0092] 6) Input all individuals in the group into the RBF model for calculation. The parameter combination with the highest predicted lift-to-drag ratio will return to step 2, perform CFD calculation on the airfoil it represents, and start the next loop. The loop will stop when a preset condition is met. In this example, the stopping condition is that the CFD evaluation reaches 300 times.

[0093] In the numerical experiments, this invention uses the NACA0012 airfoil as a benchmark and employs the CST parameterization method for 12-dimensional parameter representation. The airfoil optimization index is the lift-to-drag ratio L / D, and the CFD calculation conditions are an incoming flow speed of Mach 0.5, an airfoil angle of attack of 2°, and a Reynolds number Re = 5.0 × e 6In the algorithm, the population size s = 50, meaning the initial sampling is also 50, and the maximum number of CFD evaluations K = 300. To fully demonstrate the effectiveness of this invention, a comparative experiment was conducted. First, the airfoil optimization method provided by this invention is denoted as IA-SAHEO, and compared with two other optimization methods, SA-MPSO and SAMSO. After 10 independent computer experiments, the optimization results were statistically analyzed. The data represents the lift-to-drag ratio, and the results are shown in Table 1.

[0094] Table 1: Comparison Results of Airfoil Optimization Design

[0095] SAMSO 1.15E+02 1.06E+02 9.84E+01 6.11E+00 SA-MPSO 1.18E+02 1.08E+02 9.90E+01 5.94E+00 IA-SAHEO 1.22E+02 1.13E+02 9.88E+01 6.62E+00

[0096] As can be seen from Table 1, within the design space, the present invention provides the optimal airfoil optimization results. Figure 3 The lift-to-drag ratio convergence plots of the present invention and two comparative algorithms are presented during the optimization process. It can be seen that the airfoil optimization design method provided by the present invention is superior to the comparative algorithms throughout the entire process. Figure 4 The design results of the airfoil are shown.

[0097] In summary, this invention provides a method and system for airfoil lift-to-drag ratio optimization based on design space excitation allocation. The method includes: parameterizing the airfoil shape to be optimized and obtaining the descriptive form of the airfoil optimization problem; selecting sample points in the airfoil design space and calculating the lift-to-drag ratio of the airfoil they describe, establishing a preliminary global radial basis function (RBF) interpolation model and a local radial basis function (RBF) interpolation model from design variables to target variables; using a teaching and learning optimization method to drive the global RBF model for optimization, and using a sociological particle swarm optimization method to drive the local RBF model for optimization; using a novel design space excitation allocation strategy to drive the two optimizers to work in coordination, generating a pre-selected population in the design space; using a complex pre-selection strategy to update the pre-selected points, supplementing samples after calculation, and then updating the model until the convergence condition is met and the optimization ends. This invention determines airfoil design parameters by exploring and developing a design parameter space that can represent the airfoil shape, saving simulation evaluation times and ensuring optimization quality while reducing computational costs.

[0098] Based on the above concept, this invention also provides an airfoil lift-to-drag ratio optimization system based on design space excitation allocation, including an optimization problem construction module, a model construction module, a driving module, and an optimization module; the optimization problem construction module is used to parameterize the reference airfoil shape to be optimized, obtain the mathematical description form of the design space of the reference airfoil shape, and construct the description form of the airfoil optimization problem.

[0099] The model building module selects sample points in the design space based on the description form, and performs CFD calculations on the airfoil described by the sample points to obtain the lift-to-drag ratio, and establishes a preliminary global radial basis interpolation model and a local radial basis interpolation model from the design variables to the target variables.

[0100] The driving module utilizes two optimizers to work on the global radial basis interpolation model and the local radial basis interpolation model to generate a pre-selected population in the design space. The two optimizers are driven by the incentive allocation strategy of the design space.

[0101] The optimization module uses a complex pre-selection strategy to update the pre-selection population, performs CFD calculations, and then updates the global radial basis interpolation model and the local radial basis interpolation model until the results converge, ending the optimization and obtaining the optimized airfoil lift-to-drag ratio.

[0102] The present invention can also provide a computer device, including a processor and a memory, wherein the memory is used to store a computer executable program, the processor reads the computer executable program from the memory and executes it, and the processor can implement the airfoil lift-to-drag ratio optimization method based on design space excitation allocation described in the present invention when executing the computer executable program.

[0103] On the other hand, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can implement the airfoil lift-to-drag ratio optimization method based on design space excitation allocation described in the present invention.

[0104] The computer device may be a laptop, a desktop computer, or a workstation.

[0105] The processor described in this invention may be a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), or an off-the-shelf programmable gate array (FPGA).

[0106] The memory described in this invention can be an internal storage unit of a laptop, desktop computer, or workstation, such as memory or hard disk; or it can be an external storage unit, such as a portable hard disk or flash memory card.

[0107] Computer-readable storage media can include computer storage media and communication media. Computer storage media includes volatile and non-volatile, removable and non-removable media implemented using any method or technology for storing information such as computer-readable instructions, data structures, program modules, or other data. Computer-readable storage media can include: read-only memory (ROM), random access memory (RAM), solid-state drives (SSDs), or optical discs, etc. Random access memory can include resistive random access memory (ReRAM) and dynamic random access memory (DRAM).

[0108] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for optimizing the lift-to-drag ratio of an airfoil based on design space excitation allocation, characterized in that: The airfoil aerodynamic characteristic optimization method includes the following steps: S1. Parameterize the reference airfoil shape that needs to be optimized, obtain the mathematical description of the design variable space of the reference airfoil shape, and construct the description of the airfoil optimization problem. S2, Select sample points in the design variable space based on the description form, and perform CFD calculations on the airfoil described by the sample points to obtain the lift-to-drag ratio, and establish a preliminary global radial basis interpolation model and a local radial basis interpolation model from the design variables to the target variables. S3 utilizes two optimizers operating on a global radial basis function (RBF) interpolation model and a local radial basis function (DRB) interpolation model to generate a pre-selected population in the design space. Both optimizers are driven by an incentive allocation strategy within the design space. In S3, the first optimizer is a global RRB interpolation optimization driven by a teaching-learning optimization algorithm, where the global RRB interpolation model is constructed from all samples in the sample set. The second optimizer is a local RRB interpolation optimization driven by a social learning particle swarm optimization algorithm, where the local RRB interpolation model is built from a subset of superior samples in the sample set, and its range is limited. When one optimizer operates on either the global or local RRB interpolation model, the final pre-selected population is obtained after M iterations, during which the other optimizer remains idle. S4: Update the pre-selection population using a complex pre-selection strategy, generate supplementary samples, and perform CFD calculations on the supplementary samples. Then update the global radial basis interpolation model and the local radial basis interpolation model until the results converge, ending the optimization and obtaining the optimized airfoil lift-to-drag ratio. For the pre-selection population generated by the optimizer, implement a dimension-based permutation strategy to expand the pre-selection pool, and use the current optimizer's global radial basis interpolation model and local radial basis interpolation model to calculate all points in the pre-selection pool. Select the parameter combination corresponding to the point with the minimum calculated value for airfoil CFD calculation, with the calculation conditions being the initial conditions. Store the design parameters and calculation results in the sample set, and update the corresponding evaluation scores based on the calculation results. Then return to S3 until the calculation results reach the convergence condition.

2. The airfoil lift-to-drag ratio optimization method based on design space excitation allocation according to claim 1, characterized in that, In S1, for the reference airfoil that needs optimization, the CST parameterization method is used to parameterize the airfoil surface. d CST parameters are used as variables for airfoil optimization design, thus determining the design variables as follows: ,in For the i-th dimension of the design variable: And obtain the formulation of the airfoil lift-to-drag ratio optimization problem: in, These represent the drag coefficient and lift coefficient, respectively. These represent the baseline airfoil thickness and the optimized airfoil thickness, respectively, in the second constraint. This represents the design variable value for the i-th dimension corresponding to the reference airfoil.

3. The airfoil lift-to-drag ratio optimization method based on design space excitation allocation according to claim 1, characterized in that, In S2, sample points in the design space are selected based on the mathematical description, and CFD calculations are performed on the airfoil described by the sample points to obtain the lift-to-drag ratio. When obtaining the model from the initial design variables to the target variables, Latin hypercube design is used to extract s samples in the design space: CFD calculations were performed on the airfoil it represents. Assuming initial conditions, the calculated target variable—the ratio of drag to lift—is: The design variables and the corresponding target variables form a sample set, and a global radial basis interpolation model for the sample set is established using radial basis interpolation with a cubic function as the kernel function.

4. The airfoil lift-to-drag ratio optimization method based on design space excitation allocation according to claim 1, characterized in that, In S3, the coordinated work of the two optimizers is driven by the incentive allocation strategy in the design space. The incentive allocation strategy scores the optimizers after each new sampling and selects the optimizer with the higher score in the next round of work.

5. The airfoil lift-to-drag ratio optimization method based on design space excitation allocation according to claim 4, characterized in that, In S3, the incentive allocation strategy first designs the contribution of new sampling to exploration and development, and designs the evaluation score of the optimizer based on the contributions of both aspects. After every M iterations, the evaluation score of the working optimizer is updated, but the evaluation score of the idle optimizer is not updated.

6. An airfoil lift-to-drag ratio optimization system based on design space excitation allocation, characterized in that, The method for optimizing the lift-to-drag ratio of an airfoil based on design space excitation allocation as described in any one of claims 1-5 includes an optimization problem construction module, a model construction module, a driving module, and an optimization module. The optimization problem construction module is used to parameterize the reference airfoil shape that needs to be optimized, obtain the mathematical description of the design space of the reference airfoil shape, and construct the description of the airfoil optimization problem. The model building module selects sample points in the design space based on the description form, and performs CFD calculations on the airfoil described by the sample points to obtain the lift-to-drag ratio, and establishes a preliminary global radial basis interpolation model and a local radial basis interpolation model from the design variables to the target variables. The driving module utilizes two optimizers to work on the global radial basis interpolation model and the local radial basis interpolation model to generate a pre-selected population in the design space. The two optimizers are driven by the incentive allocation strategy of the design space. The optimization module uses a complex pre-selection strategy to update the pre-selection population, performs CFD calculations, and then updates the global radial basis interpolation model and the local radial basis interpolation model until the results converge, ending the optimization and obtaining the optimized airfoil lift-to-drag ratio.

7. A computer device, characterized in that, It includes a processor and a memory, the memory being used to store a computer-executable program, the processor reading the computer-executable program from the memory and executing it, and the processor executing the computer-executable program being able to implement the airfoil lift-to-drag ratio optimization method based on design space excitation allocation as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, A computer-readable storage medium stores a computer program that, when executed by a processor, enables the airfoil lift-to-drag ratio optimization method based on design space excitation allocation as described in any one of claims 1 to 5.

Images