Aerodynamic shape design method based on global intelligent optimization algorithm

Abstract

The application provides a kind of aerodynamic shape design method based on global intelligent optimization algorithm, first according to engineering demand to determine the value range of design variable, generates initial sample point set through Latin hypercube sampling and other experimental design methods, obtains the aerodynamic performance data corresponding to sample point based on high-precision CFD numerical simulation;Subsequently, the data set is used to construct a Bayesian neural network proxy model, which can not only realize the efficient prediction of aerodynamic performance, but also quantify the prediction confidence of the model through the output prediction variance;In the optimization iteration process, the prediction uncertainty is guided in the form of penalty function aerodynamic optimization, and TLBO global optimization algorithm is used to search for the optimal solution in the design space that meets multiple constraint conditions;By introducing the "conservative factor" to dynamically control the weight coefficient of the penalty function, the adaptive balance of global exploration and local optimization is realized, and finally the optimal aerodynamic shape scheme that meets the engineering design requirements is output.

Description

Technical Field

[0001] This invention belongs to the field of intelligent algorithms and is used for the aerodynamic layout optimization design of aircraft. Specifically, it relates to an aerodynamic shape design method based on a global intelligent optimization algorithm. Background Technology

[0002] In the field of aircraft aerodynamic shape optimization, traditional optimization methods generally suffer from high computational costs, long design cycles, and a tendency to get trapped in local optima, making it difficult to meet the needs of rapid iteration and global optimization. Existing multi-objective surrogate model techniques have significant deficiencies in accuracy and generalization ability, especially in highly nonlinear problems, resulting in large deviations between optimization results and actual data, thus limiting their application in engineering. Furthermore, when artificial intelligence technology is introduced into aerodynamic optimization, the "illusion" problem (i.e., model predictions do not match physical properties) has not been effectively resolved, potentially leading to erroneous design decisions and jeopardizing project safety. Although surrogate models are widely used in optimization, existing methods still struggle to balance efficiency, globality, and accuracy. Summary of the Invention

[0003] To address the aforementioned issues, this invention proposes an aerodynamic shape design method based on a global intelligent optimization algorithm. Through intelligent modeling and dynamic autonomous decision-making, it breaks through the bottlenecks of traditional optimization methods, achieving coordinated development of global search capability and local efficiency in multi-objective optimization. It also improves the prediction accuracy and generalization ability of the surrogate model, avoids design risks caused by the "illusion" problem, and significantly enhances the efficiency and stability of aerodynamic shape optimization design.

[0004] An aerodynamic shape design method based on a global intelligent optimization algorithm includes the following steps: (1) Determine the range of design variables based on engineering requirements, generate an initial sample point set through experimental design methods such as Latin hypercube sampling, and obtain the aerodynamic performance data corresponding to the sample points based on high-precision CFD numerical simulation. (2) Construct a Bayesian neural network agent model using this dataset; (3) During the optimization iteration process, the prediction uncertainty is used as a penalty function to guide the aerodynamic optimization, and the TLBO global optimization algorithm is used to search for the optimal solution in the design space that satisfies multiple constraints. (4) By introducing a “conservative factor” to dynamically adjust the weight coefficient of the penalty function, an adaptive balance between global exploration and local optimization is achieved, and the optimal aerodynamic shape scheme that meets the engineering design requirements is finally output.

[0005] The specific steps of the TLBO global optimization algorithm are as follows: Step 1: Set initial parameters and maximum number of iterations ; Step 2: Population initialization and evaluation; Step 3: Select the best individual as the teacher And calculate the average knowledge level ; Step 4: Perform the "Teaching" phase; Step 5: Implement the "Learning" phase; Step 6: Check if the convergence condition has been met. If it is met, end the optimization. If not, increase the number of iterations G and return to step 3 to continue.

[0006] In step 4, during the teaching phase, the algorithm first determines the best individual in the current population as the "teacher," and then updates the population through the following steps: in, It is the average solution of the population. This is the updated student solution. This is the current student's solution. He is a teacher. It is a teaching factor that determines the strength of a teacher's influence on students; it can be 1 or 2, and is usually randomly selected. yes A random number within a given range.

[0007] In step 5, during the learning phase, students improve their skills by learning from each other; each student Another student will be randomly selected. To learn, if Compare If it is even better, then: if Compare If it is better, then: in This is the updated student solution. and These are the current solutions for the two selected students.

[0008] The beneficial effects of this invention are as follows: This invention effectively avoids the "curse of dimensionality" problem of proxy models and improves the efficiency of sample configuration in high-dimensional design spaces; it suppresses the risk of "illusion" through a dynamic learning mechanism and significantly reduces design errors caused by model bias; compared with traditional particle swarm optimization or genetic algorithms, this invention can achieve balanced optimization results without cumbersome parameter settings, saving design and development investment and shortening the decision-making cycle. Attached Figure Description

[0009] Figure 1 This is a flowchart of an aerodynamic shape design method based on a global intelligent optimization algorithm. Detailed Implementation

[0010] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection claimed by the present invention.

[0011] An aerodynamic shape design method based on a global intelligent optimization algorithm includes the following steps: First, the range of design variables is determined based on engineering requirements. An initial sample point set is generated using experimental design methods such as Latin hypercube sampling. Aerodynamic performance data corresponding to the sample points is obtained based on high-precision CFD numerical simulation. Then, a Bayesian neural network surrogate model is constructed using this dataset. This model can not only achieve efficient prediction of aerodynamic performance, but also quantify the prediction confidence of the model by outputting the prediction variance. During the optimization iteration process, the prediction uncertainty is used as a penalty function to guide aerodynamic optimization. The TLBO global optimization algorithm is used to search for the optimal solution in the design space that meets multiple constraints. By introducing a "conservative factor" to dynamically adjust the weight coefficient of the penalty function, an adaptive balance between global exploration and local optimization is achieved, and finally, the optimal aerodynamic shape scheme that meets the engineering design requirements is output.

[0012] The mathematical representation of the aerodynamic optimization model is as follows: In the above global optimization model, This represents an n-dimensional design variable vector, whose value boundary is determined by... (lower limit) and (Upper limit) definition; To optimize the objective function, key aerodynamic performance indicators such as the aircraft's drag coefficient and lift-to-drag ratio can be selected, with the optimization objective being to minimize the function value by adjusting the design variables. Let represent the i-th inequality constraint function, and m be the total number of constraints. These constraints cover practical engineering requirements such as geometric feasibility restrictions and lower limits of aerodynamic performance indicators, ensuring that the optimization process always proceeds within the feasible region. Predicted uncertainties are difficult to accurately match actual uncertainties.

[0013] Under high-dimensional, small-sample conditions, the prediction uncertainty of surrogate models is difficult to precisely match with the uncertainty of real physical processes. The main reasons for this are twofold: first, limited sample data cannot fully characterize the multi-scale nonlinear characteristics of complex three-dimensional flows, leading to systematic biases in model predictions; second, the mathematical assumptions of traditional surrogate models differ fundamentally from the strong coupling effect of actual flow phenomena, further amplifying prediction uncertainty. To address this issue, this paper uses the prediction variance output by a Bayesian neural network to quantify uncertainty. The magnitude of the prediction variance directly characterizes the level of model prediction uncertainty. This quantification method can accurately identify low-reliability prediction regions during the optimization process, effectively avoiding the risk of blind searching.

[0014] The objective function of the original aerodynamic optimization problem is as follows: The currently developed method improves the objective function as follows: in To comprehensively consider the loss function under constraints, The original objective function is... This is the penalty factor parameter, used to adjust the intensity of the penalty for constraint violation; To constrain the calculation of violation items; It is a penalty function term that integrates prediction uncertainty. Its function construction comprehensively considers multiple dimensions of information, including model prediction variance, constraint violation degree, and objective function value. Conservative factor. The sign and magnitude of the value determine the orientation of the optimization strategy: when When the value is positive, the penalty for constraint violation increases, prompting the algorithm to perform a refined search within the current high-confidence region, exhibiting local optimization characteristics; when When negative values ​​are taken, the penalty mechanism transforms into an exploration incentive, guiding the algorithm to expand into potential optimization regions outside the constraint boundaries, thus achieving a global search function. In the actual optimization process, a model is constructed based on characteristic parameters such as the optimization iteration algebra number and the rate of change of the mean variance of the model prediction. The adaptive update strategy dynamically adjusts parameter values ​​as the confidence of the surrogate model increases, achieving a synergistic improvement in optimization efficiency and quality finding while ensuring constraint satisfaction.

[0015] The TLBO algorithm likens the optimization problem to a teaching and learning process. In this mechanism, the "teacher" represents the current optimal solution, and the "student" simulates the learning process by moving towards the teacher's position. The moving distance is dynamically adjusted based on the differences between the teacher and the student. The algorithm consists of two core phases: First, in the teacher's teaching phase, students move closer to the optimal solution to enhance global search capabilities; second, in the student learning phase, group collaboration breaks through the limitations of local optima, significantly improving convergence efficiency and stability.

[0016] The TLBO algorithm mainly consists of two phases: the teacher learning phase and the student learning phase. In the teacher learning phase, the algorithm selects the best solution in the current population as the teacher. Other students learn by moving towards the teacher's position; the movement distance is adjusted based on the difference between the teacher and the students. The specific steps of the TLBO algorithm are as follows:

[0017] Step 1: Set initial parameters and maximum number of iterations .

[0018] Step 2: Population initialization and evaluation.

[0019] Step 3: Select the best individual as the teacher And calculate the average knowledge level .

[0020] Step 4: Perform the "Teaching" Phase. In the teaching phase, the algorithm first determines the best individual in the current population as the "teacher." Then, the algorithm updates the population through the following steps:

[0021] in, It is the average solution of the population. This is the updated student solution. This is the current student's solution. He is a teacher. It is a teaching factor that determines the strength of a teacher's influence on students; it can be 1 or 2, and is usually chosen randomly. yes A random number within a given range.

[0022] Step 5: Implement the "Learning" Phase. In the learning phase, students improve their skills by learning from each other. Each student... Another student will be randomly selected. To learn, if Compare If it is even better, then:

[0023] if Compare If it is better, then: in This is the updated student solution. and These are the current solutions for the two selected students.

[0024] Step 6: Check if the convergence condition has been met. If it is met, end the optimization. If not, increase the number of iterations G and return to step 3 to continue.

[0025] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An aerodynamic shape design method based on a global intelligent optimization algorithm, characterized in that, Includes the following steps: (1) Determine the range of design variables based on engineering requirements, generate an initial sample point set through experimental design methods such as Latin hypercube sampling, and obtain the aerodynamic performance data corresponding to the sample points based on high-precision CFD numerical simulation. (2) Construct a Bayesian neural network agent model using this dataset; (3) During the optimization iteration process, the prediction uncertainty is used as a penalty function to guide the aerodynamic optimization, and the TLBO global optimization algorithm is used to search for the optimal solution in the design space that satisfies multiple constraints. (4) By introducing a “conservative factor” to dynamically adjust the weight coefficient of the penalty function, an adaptive balance between global exploration and local optimization is achieved, and the optimal aerodynamic shape scheme that meets the engineering design requirements is finally output.

2. The aerodynamic shape design method based on a global intelligent optimization algorithm according to claim 1, characterized in that, The specific steps of the TLBO global optimization algorithm are as follows: Step 1: Set initial parameters and maximum number of iterations ; Step 2: Population initialization and evaluation; Step 3: Select the best individual as the teacher And calculate the average knowledge level ; Step 4: Perform the "Teaching" phase; Step 5: Perform the "Learning" phase; Step 6: Check if the convergence condition has been met. If it is met, end the optimization. If not, increase the number of iterations G and return to step 3 to continue.

3. The aerodynamic shape design method based on a global intelligent optimization algorithm according to claim 2, characterized in that, In step 4, during the teaching phase, the algorithm first determines the best individual in the current population as the "teacher," and then updates the population through the following steps: in, It is the average solution of the population. This is the updated student solution. This is the current student's solution. He is a teacher. It is a teaching factor that determines the strength of a teacher's influence on students; it can be 1 or 2, and is usually randomly selected. yes A random number within a given range.

4. The aerodynamic shape design method based on a global intelligent optimization algorithm according to claim 2, characterized in that, In step 5, during the learning phase, students improve their skills by learning from each other; each student Another student will be randomly selected. To learn, if Compare If it is even better, then: if Compare If it is better, then: in This is the updated student solution. and These are the current solutions for the two selected students.

Images