Propeller airfoil aerodynamic optimization method based on DBO-BP and NSWOA

Through the combination of DBO-BP and NSWOA algorithms, the aerodynamic performance of the propeller airfoil is optimized, which solves the problems of large amount of calculation and poor fitting effect of the agent model in the prior art, and achieves more efficient aerodynamic optimization and reliability improvement.

CN118965577APending Publication Date: 2024-11-15CHINA JILIANG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411070416.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-06
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

When dealing with propeller airplane aerodynamic optimization, the calculation amount is large, the calculation time is long, and when nonlinear problems or design variables are large, the agent model fitting effect is poor, resulting in low reliability and unsatisfactory optimization effect.

Method used

Adopting propeller wing aerodynamic optimization method based on DBO-BP and NSWOA, the weight and threshold of the BP neural network are optimized through the beetle optimization algorithm, and multi-objective optimization of lift coefficient and drag coefficient is combined with the NSWOA algorithm to improve the accuracy and optimization of the agent model. Reliability of results.

Benefits of technology

It effectively improves the accuracy of the BP neural network proxy model and the reliability of the optimization results, solves the problems of single optimization goals and low optimization efficiency, and significantly improves the aerodynamic performance of the propeller wing.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118965577A_ABST
    Figure CN118965577A_ABST
Patent Text Reader

Abstract

The invention discloses a propeller airfoil aerodynamic optimization method based on DBO-BP and NSWOA, and belongs to the technical field of aircrafts. The method comprises the following steps: firstly, parameterizing an original airfoil profile through a CST airfoil profile parametric method, and disturbing the original airfoil profile within a certain range by combining a Latin hypercube sampling method to generate a sample space; thirdly, performing pneumatic analysis on the airfoil profile in the sample space by using CFD software, and constructing a sample data set; then, a DBO algorithm is used for optimizing the initial weight and the threshold value of the BP proxy model, and establishment of a DBO-BP neural network proxy model is achieved; and finally, optimizing the agent model in combination with an NSWOA algorithm, and verifying and analyzing an optimization result through CFD software. According to the method, the DBO-BP proxy model is constructed, so that the reliability of the BP proxy model and an optimization result can be effectively improved; through the NSWOA algorithm, the lift coefficient and the resistance coefficient are optimized at the same time, and the problems that the optimization target is single and the optimization efficiency is low can be effectively solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of aircraft, and in particular is a propeller airfoil aerodynamic optimization method based on DBO-BP and NSWOA. Technical Background

[0002] Propeller aircraft are widely used in military and civilian scenarios due to their economical and adaptable characteristics, playing an irreplaceable role in the aviation field. With the needs of tactical transportation, the requirements for the flight performance of propeller UAVs are also constantly increasing. Under this development requirement, the aerodynamic optimization of propellers is still one of the key issues. In the aerodynamic optimization of propellers, the airfoil, as the basis of the blades, fundamentally determines the performance of the blades. Only by selecting the appropriate airfoil can the efficiency of the propeller be maximized.

[0003] In order to solve the problem of excessive calculation or long calculation time in complex engineering design in the aerospace field, and to obtain the highest performance design results, an optimization method based on surrogate model has emerged. However, when dealing with strong nonlinear problems or many design variables, the polynomial response surface fitting effect is poor, resulting in low reliability of the surrogate model and unsatisfactory optimization effect.

[0004] To this end, the present invention proposes a propeller airfoil aerodynamic optimization method based on DBO-BP and NSWOA. Summary of the invention

[0005] In order to overcome the shortcomings of the prior art, the present invention provides a propeller airfoil aerodynamic optimization method based on DBO-BP (Dung Beetle Optimization-Back Propagation) and NSWOA (Non-Dominated Sorting Whale Optimization Algorithm), which can effectively improve the accuracy of the BP neural network proxy model and improve the reliability of the constructed proxy model and the optimization results; combining the NSWOA algorithm to optimize the lift coefficient and the drag coefficient can effectively solve the problems of single optimization target and low optimization efficiency.

[0006] Specifically, the present invention provides a propeller airfoil aerodynamic optimization method based on DBO-BP and NSWOA, the method comprising:

[0007] Step 1: Determine the initial airfoil and parameterize it using the CST parameter method to obtain the original CST parameters.

[0008] Step 2: Combined with the Latin hypercube sampling method, uniform random sampling is performed according to the original CST parameter control points within a reasonable range to generate the sample space.

[0009] Step 3: Combined with the CST parameter method, the airfoil parameter control points in the sample space are converted into airfoil coordinate points.

[0010] Step 4: Draw the mesh of the original airfoil to obtain the initial mesh structure.

[0011] Step 5: Call ICEM through the script to draw the mesh of all airfoils within the sample range.

[0012] Step 6: Call Fluent through the script to perform aerodynamic analysis on the generated mesh and output the obtained Cl and Cd files.

[0013] Step 7: Combine the CST parameter set with the corresponding simulation results to create a sample data set.

[0014] Step 8: Normalize the samples.

[0015] Step 9: Shuffle the sample data set and divide it into training set and test set according to a certain ratio.

[0016] Step 10: Establish the initial BP neural network and use the multi-layer perceptron neural network to build the neural network.

[0017] Step 11: Map the weights and thresholds of the initialized BP neural network to the dung beetle optimization algorithm for optimization, and output the optimal weights and thresholds after optimization by the dung beetle optimization algorithm.

[0018] Step 12: When the requirements are met, that is, the algorithm reaches the maximum number of iterations, the optimal weight and threshold of the original BP neural network are output.

[0019] Step 13: Combine the BP neural network to continuously perform training and testing until the termination condition is met, that is, the proxy model is established.

[0020] Step 14: Combine the NSWOA optimization algorithm to perform multi-objective optimization on the airfoil parameters.

[0021] Step 15: Use CFD methods to verify the accuracy of the proxy model prediction data and the reliability and effectiveness of the optimization results.

[0022] Preferably, the specific formula for parameterizing the CST airfoil in step 1 is as follows:

[0023]

[0024]

[0025] In the above formula: (ζ) Upper is the formula for the upper surface of the airfoil, (ζ) Loweris the formula for the lower surface of the airfoil. ψ=x / c, ζ=y / c are the dimensionless coordinates of the airfoil; Δξ Upper =zu TE / C is the leading edge thickness ratio of the upper surface of the airfoil; Δξ Lower =zl TE / C is the thickness ratio of the trailing edge of the lower surface of the airfoil.

[0026] Preferably, when the mesh is drawn in step 4, mesh independence verification should be performed to eliminate the influence of the number of meshes on the airfoil simulation results.

[0027] Preferably, in step 5, the grid structure and quantity of all airfoils should be consistent with the original airfoil grid.

[0028] Preferably, in step 6, the flow field solution of the airfoil is based on the RANS equation of the SST turbulence model, the time term adopts a second-order implicit format, the convection term is discretized using a second-order headwind format, and the diffusion term adopts a central format.

[0029] Preferably, in step 6, the set incoming flow angle of attack, incoming flow velocity and pressure should be ensured to remain consistent during the CFD simulation process.

[0030] Preferably, in step 8, the Map minmax function provided by MATLAB is selected for normalization and denormalization, as shown in the following formula:

[0031]

[0032] In the above formula, Y max and Y min are the maximum and minimum values ​​of the samples, respectively; X and Y are the values ​​before and after the transformation, respectively.

[0033] Preferably, in step 10, the number of neurons in the input layer is determined by the number of design variables, the number of neurons in the output layer is determined by the number of optimization targets, and the number of neurons in the hidden layer is determined according to an empirical formula, which is as follows:

[0034]

[0035] In the above formula, L is the number of hidden layer nodes, a is the number of input nodes, b is the number of output nodes, and c is a constant between [1,10].

[0036] Preferably, in step 11, the steps of optimizing the initial weights and thresholds of the BP neural network using the dung beetle algorithm are as follows:

[0037] Step 11.1 initialize the parameters of the dung beetle optimization algorithm according to the number of neurons in the input layer, output layer and hidden layer of the BP neural network;

[0038] Step 11.2 When the dung beetle moves forward, if there is no obstacle, it updates its position by rolling the ball. If there is an obstacle, it updates its position by dancing. The position update formula is as follows:

[0039]

[0040] In the above formula, k is the current iteration coefficient; is the position of the i-th dung beetle at the k-th iteration; is the position of the i-th dung beetle at the k-1-th iteration; is the position of the i-th dung beetle at the k+1th iteration; α is the natural coefficient, which takes the value of 1 or -1, that is, when it takes 1, it does not deviate from the original direction, and when it takes -1, it deviates from the original direction; e is the deflection coefficient, which takes a constant between (0,0.2]; b takes a constant between (0,1); X ω is the global worst position; θ is the deflection angle, which is an angle between [0,π]. When θ is 0, π / 2 or π, the position of the dung beetle is not updated. When θ is other angles, the position of the dung beetle is updated.

[0041] Step 11.3 simulates the dung beetle egg-laying area through the boundary selection strategy, and its boundary selection is defined as follows:

[0042]

[0043] In the above formula, U * , L * are the upper and lower bounds of the boundary region; U and L are the upper and lower bounds of the optimization problem respectively; X * is the current local optimal position; R is the intermediate quantity, R = 1-k / K max , K max is the maximum number of iterations;

[0044] Step 11.4 According to the boundary selection definition, the dung beetle egg-laying area changes dynamically with the number of iterations, so the position of the egg ball also changes dynamically, which is expressed as follows:

[0045]

[0046] In the above formula, is the position of the i-th egg ball at the k-th iteration; is the position of the i-th egg ball at the k+1-th iteration; r 1 and r 2 is an independent random vector of 1×D; D is the dimension;

[0047] Step 11.5 guides the grown dung beetles to forage by establishing an optimal foraging area, and the optimal foraging boundary area is defined as follows:

[0048]

[0049] In the above formula, Ub , L b are the upper and lower bounds of the foraging area respectively; X b is the global optimal position;

[0050] Step 11.6 defines the foraging position update according to the dung beetle egg-laying position update principle, which is defined as follows:

[0051]

[0052] In the above formula, c 1 is a random number that follows a normal distribution; c 2 is a random variable between (0,1);

[0053] Step 11.7 Set X b The surroundings are the best positions to grab the dung ball, so the position update definition of the dung beetle grabbing the dung ball is as follows:

[0054]

[0055] In the above formula, s is a constant; g is a 1×D random vector that obeys a normal distribution;

[0056] Step 11.8 Select the root mean square error Z RMSE is the fitness function of the dung beetle optimization algorithm, Z RMSE is defined as follows:

[0057]

[0058] In the above formula: n is the number of samples, y is j is the actual value of the jth group of data, is the predicted value of the jth group of data.

[0059] In step 11.9, after the dung beetle optimization algorithm has been optimized to the maximum number of iterations, the initial weights and thresholds of the optimal BP neural network can be obtained.

[0060] Preferably, in step 12, mean absolute error (MAE), mean square error (MSE) and mean absolute percentage error (MAPE) are selected as the evaluation indicators. These three indicators can measure the gap between the predicted value and the actual value from different angles. MAE reflects the overall error, MSE focuses on the prediction deviation, and MAPE focuses on the prediction accuracy. The smaller these three indicators are, the higher the accuracy of the proxy model. The expressions of each error formula are:

[0061]

[0062]

[0063]

[0064] In the above formula: n is the number of samples, y is j is the actual value of the jth group of data, is the predicted value of the jth group of data.

[0065] Preferably, in step 14, the steps of optimizing airfoil parameters using the NSWOA algorithm are as follows:

[0066] Step 14.1 Initialize the parameters of the NSWOA algorithm, select the initial population, and initialize according to the constraints. Initialize the population size N, the maximum number of iterations T, and the whale position information X(t) of the NSWOA algorithm, calculate the corresponding fitness value of each whale, sort them according to the size of the fitness value, and select a population size of N whales as the initial population.

[0067] Step 14.2 updates the current optimal solution through the encirclement mechanism, spiral mechanism and foraging mechanism of the NSWOA algorithm.

[0068] Step 14.2.1 Encirclement mechanism. Since the optimal position in the search space is unknown, it is assumed that the current best candidate position is the target prey position. After the target prey position is determined, other whales will try to encircle the target prey position. The established mathematical formula is as follows:

[0069] X(t+1)=X * (t)-A·D

[0070] D=|C·X * (t)-X(t)|

[0071] In the above formula, t represents the current iteration number; A and C represent coefficients; X * (t) represents the best whale position vector so far; X(t) represents the current whale position vector. A and C can be obtained by the following formulas:

[0072]

[0073] C=2r 2

[0074] In the above formula, r 1 and r 2 is a random number in (0,1); a is the convergence factor, which decreases linearly from 2 to 0 during the iteration; t represents the current number of iterations;

[0075] Step 14.2.2 Spiral mechanism. The whale will shrink the encirclement while swimming towards the prey in a spiral shape. In the synchronous behavior model, it is assumed that there is a probability of l to select the shrinking encirclement mechanism and a probability of l to select the spiral model to update the whale's position, so the established mathematical formula is as follows:

[0076]

[0077] In the above formula, D P represents the distance between the whale and the prey, b is a constant used to define the spiral shape, l is a random number between (-1,1); p represents the probability of the predation mechanism, which is a random number between [0,1]. As the number of iterations t increases, the parameter A and the convergence factor a gradually decrease. If |A| < 1, the whales gradually surround the current optimal solution, which is the local optimization stage;

[0078] Step 14.2.3 Foraging mechanism. To ensure that all whales can fully search in space, the positions of whales are changed according to the distance between them to achieve the purpose of global random search. The established mathematical formula is as follows:

[0079] X(t+1)=X rand (t)-A·D

[0080] D=|C·X rand (t)-X(t)|

[0081] In the above formula, X rand (t) represents the position of a random individual whale in the current whale group.

[0082] Step 14.3 Repeat step 14.2 until the maximum number of iterations is reached and the Pareto optimal solution set of Cl and Cd is obtained.

[0083] Preferably, in step 14, the optimization range of the airfoil parameters should be set within a reasonable range to ensure that the airfoil will not be excessively deformed.

[0084] Preferably, in step 14, the optimization target is set to maximize lift and minimize drag.

[0085] Compared with the prior art, the present invention optimizes the BP neural network agent through the dung beetle optimization algorithm to construct a DBO-BP neural network agent model between airfoil parameters and aerodynamic performance, and then optimizes the lift coefficient and drag coefficient through the NSWOA algorithm. The DBO-BP neural network agent model constructed by this method can effectively improve the accuracy of the original BP agent model and enhance the reliability of the agent model and optimization results; through the NSWOA algorithm, the lift coefficient and drag coefficient are optimized at the same time, which can effectively solve the problem of single optimization target and low optimization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 This is the overall flow chart of this method.

[0087] Figure 2 Comparison diagram of residuals of airfoils of different orders.

[0088] Figure 3 is the airfoil diagram in the sample space.

[0089] Figure 4 Schematic diagram of O-type mesh division.

[0090] Figure 5 Schematic diagram of the neural network structure.

[0091] Figure 6 Schematic diagram of lift test results of BP proxy model.

[0092] Figure 7 Schematic diagram of the resistance test results of the BP proxy model.

[0093] Figure 8 Schematic diagram of lift test results of DBO-BP proxy model.

[0094] Fig. 9 Schematic diagram of the resistance test results of the DBO-BP proxy model.

[0095] Fig.10 Comparison diagram of airfoil before and after optimization.

[0096] Fig.11 Comparison of lift coefficients of front and rear optimized airfoils.

[0097] Fig.12 Comparison of drag coefficients of the airfoil before and after optimization. DETAILED DESCRIPTION

[0098] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below with reference to the accompanying drawings and examples.

[0099] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which the present invention belongs. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0100] A propeller airfoil aerodynamic optimization design method based on DBO-BP and NSWOA algorithms. The overall process is as follows: Figure 1 As shown, the following steps are included:

[0101] In this example, NACA0012 airfoil is used as the original airfoil for aerodynamic optimization.

[0102] 1. The airfoil selected in this method is NACA0012 airfoil, and the design conditions are: the incoming flow angle of attack is 2°, the incoming flow Mach number is 0.8, and the Reynolds number is 6×10 6 .

[0103] Step 1: CST parameterize the original airfoil, i.e., the NACA0012 airfoil, and combine Figure 2 ,Considering the number of CST parameters and the description effect, ,the third-order Bernstein polynomial is selected in this example, and 4 ,design variables are used to describe the upper and lower surfaces, ,a total of 8 design variables, which can better describe the airfoil ,with relatively few parameters.

[0104] Step 2: Combined with the Latin hypercube sampling method, uniform random sampling is performed on the initial CST parameter control points within a reasonable range to generate a sample space. The airfoil in the sample space is as follows: Figure 3 shown.

[0105] Step 3: Combined with the CST parameter method, the airfoil parameter control points in the sample space are converted into airfoil coordinate points.

[0106] Step 4: Draw the mesh of the original airfoil to obtain the initial mesh structure. Use O mesh to draw the airfoil, such as Figure 4 shown.

[0107] Step 5: Call ICEM through the script to draw the mesh of all airfoils within the sample range.

[0108] Step 6: Perform aerodynamic analysis on the generated mesh by calling Fluent through the script, and output the obtained Cl and Cd files. The Y+ value of the airfoil wall is 1, and the airfoil flow field solution is based on the RANS equation of the SST turbulence model. The time term uses the second-order implicit format, the convection term is discretized using the second-order headwind format, and the diffusion term uses the center format. The incoming flow angle, incoming flow velocity and pressure set during the CFD simulation process remain consistent.

[0109] Step 7: Combine the simulation results corresponding to the CST parameter set to create a sample data set.

[0110] Step 8: Normalize the samples.

[0111] Step 9: Shuffle the sample data set and divide it into a training set and a test set according to a certain ratio. In this example, 80% of the total samples are used as the training set, and 20% of the total samples are used as the test set.

[0112] Step 10: Establish the initial BP network and use the multi-layer perceptron neural network to build the neural network, such as Figure 5 As shown in the figure, the network contains an input layer, a hidden layer, and an output layer. The input layer contains 8 neurons, and the input is 8 CST parameters; the hidden layer has 1 layer, and the number of neurons is 14; the output layer has 2 neurons, and the output is the CFD simulation results Cl and Cd of the CST parameterized airfoil.

[0113] Step 11: After initializing the dung beetle population, the weights and thresholds of the initialized BP neural network are mapped to the dung beetle optimization algorithm for optimization. After optimization by the dung beetle optimization algorithm, the optimal weights and thresholds are output. The parameters in the dung beetle algorithm are set as follows: the total number of dung beetles is 30, the maximum number of iterations is 50, the obstacle avoidance probability is 0.1, the deflection coefficient is 0.5, and the constants b and S are 0.3 and 0.5 respectively.

[0114] Step 12: When the algorithm reaches the maximum number of iterations, it meets the requirements to output the optimal weights and thresholds of the original BP neural network.

[0115] Step 13: Combine the BP neural network to continuously train and test, and set the optimization goal to minimize the loss function until the loss function of the training sample training set no longer decreases, that is, the training is completed, and then combined with the test set for testing. The lift and drag test results of the BP proxy model and the DBO-BP proxy model can be obtained, as shown in Figure 6 , Figure 7 , Figure 8 , Fig. 9 As shown. Then, the accuracy of the DBO-BP proxy model and the BP proxy model is compared by comparing their MSE, MAE, and MAPE. As shown in Tables 1 and 2, it can be seen that the DBO algorithm can effectively improve the accuracy of the BP proxy model.

[0116] Table 1 Comparison of lift training and test errors of BP proxy model and DBO-BP proxy model

[0117] Table 2 Comparison of resistance training and test errors between the BP surrogate model and the DBO-BP surrogate model

[0118] Step 14: Combine the constructed DBO-BP proxy model and the NSWOA multi-objective whale optimization algorithm to perform multi-objective optimization on the airfoil parameters in the data set. The optimization variables are the airfoil parameters, and the optimization objectives are the maximum lift coefficient and the minimum drag coefficient. The maximum number of iterations is set to 100, and the maximum number of populations is 50. The optimization range of the airfoil parameters should be set within a reasonable range. The optimization variable setting range of this example is shown in Table 3:

[0119] Table 3 CST parameter optimization boundary

[0120] Step 15: The CFD method is used to further verify the accuracy of the predicted data of the constructed DBO-BP proxy model and the reliability of the optimization results. As shown in the data in Table 4, the predicted value of the proxy model is slightly different from the actual value of the simulation, which meets the expected requirements. The optimized airfoil parameters are converted into coordinates by combining the CST parameter method. The airfoil is simulated by CFD under the same grid structure, boundary conditions and incoming flow conditions. The aerodynamic performance of the airfoil before and after optimization is compared. Fig.11 , Fig.12 As shown in Table 5, the converged results show that after optimization using this method, the lift coefficient, drag coefficient and lift-drag coefficient of the original airfoil are optimized to a large extent.

[0121] Table 4 Error comparison between the proxy model prediction value and the simulation actual value

[0122] Table 5 Comparison of aerodynamic performance before and after optimization

[0123] In summary, it can be seen that this method can effectively improve the reliability of the BP neural network proxy model and the aerodynamic performance of the original airfoil, which is of great help to the engineering design of propellers and various airfoils.

[0124] The above implementation modes are only for illustrating the technical concept and features of the present invention, and their purpose is to enable technicians in this field to understand the content of the present invention and implement it. They cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made according to the spirit of the present invention should be included in the protection scope of the present invention.

Claims

1. A propeller airfoil aerodynamic optimization method based on DBO-BP and NSWOA, characterized in that: The following steps are involved: Step 1: Determine the initial airfoil and parameterize it using the CST parameter method to obtain the original CST parameters; Step 2: Combined with the Latin hypercube sampling method, uniform random sampling is performed according to the original CST parameter control points within a reasonable range to generate the sample space; Step 3: Combined with the CST parameter method, the airfoil parameter control points in the sample space are converted into airfoil coordinate points; Step 4: Draw the mesh of the original airfoil to obtain the initial mesh structure; Step 5: Call ICEM through the script to draw the mesh of all airfoils within the sample range; Step 6: Call Fluent through the script to perform aerodynamic analysis on the generated Mesh grid and output the obtained Cl and Cd files; Step 7: Merge the CST parameter set with the corresponding simulation results to create a sample data set; Step 8: Normalize the samples; Step 9: Shuffle the sample data set and divide it into training set and test set according to a certain ratio; Step 10: Establish an initial BP neural network and use a multi-layer perceptron neural network to build a neural network; Step 11: Map the weights and thresholds of the initialized BP neural network to the dung beetle optimization algorithm for optimization, and output the optimal weights and thresholds after optimization by the dung beetle optimization algorithm; Step 12: When the requirements are met, that is, the algorithm reaches the maximum number of iterations, the optimal weight and threshold of the original BP neural network are output; Step 13: Combine the BP neural network to continuously train and test until the termination condition is met, that is, the agent model is established; Step 14: Combine the NSWOA optimization algorithm to perform multi-objective optimization on the airfoil parameters; Step 15: Use CFD methods to verify the accuracy of the proxy model's prediction data and the reliability of the optimization results.

2. The propeller airfoil aerodynamic optimization method based on DBO-BP and NSWOA according to claim 1, characterized in that In step 8, the Map minmax function provided by MATLAB is selected for normalization and denormalization, as shown in the following formula: In the above formula, Y max and Y min are the maximum and minimum values ​​of the samples, respectively; X and Y are the values ​​before and after the transformation, respectively.

3. The propeller airfoil aerodynamic optimization method based on DBO-BP and NSWOA according to claim 1, characterized in that In step 10, the number of neurons in the input layer is determined by the number of design variables, the number of neurons in the output layer is determined by the number of optimization targets, and the number of neurons in the hidden layer is determined according to an empirical formula, which is as follows: In the above formula, L is the number of hidden layer nodes, a is the number of input nodes, b is the number of output nodes, and c is a constant between [1,10].

4. The propeller airfoil aerodynamic optimization method based on DBO-BP and NSWOA according to claim 1, characterized in that In step 11, the initial weight and threshold of the BP neural network are optimized by using the dung beetle algorithm.

5. The propeller airfoil aerodynamic optimization method based on DBO-BP and NSWOA according to claim 1, characterized in that In step 12, mean absolute error (MAE), mean square error (MSE) and mean absolute percentage error (MAPE) are selected as evaluation indicators.

6. The propeller airfoil aerodynamic optimization method based on DBO-BP and NSWOA according to claim 1, characterized in that In step 14, the optimization range of the airfoil parameters should be set within a reasonable range to ensure that the airfoil will not be excessively deformed; the optimization target is set to maximize the lift coefficient and minimize the drag coefficient.