A hierarchical aerodynamic optimization method for micro solar-powered unmanned aerial vehicles

Abstract

This invention belongs to the field of aerodynamic optimization technology for unmanned aerial vehicles (UAVs), specifically relating to a hierarchical aerodynamic optimization method for a micro solar-powered UAV. This optimization method employs a hierarchical optimization approach, involving a two-step optimization process. First, a genetic algorithm is used to select the optimal population and optimal sample. Second, adjoint solution is performed on the current optimal sample to obtain gradient information for free deformation optimization based on mesh deformation. Furthermore, this invention employs a step-by-step optimization method, first optimizing the two-dimensional airfoil, and then optimizing the three-dimensional airfoil according to specific practical requirements. This aerodynamic optimization method does not rely on a fixed configuration, reducing the number of gradient deformation iterations and thus minimizing optimization overhead. Simultaneously, it can maximize the coverage of design conditions and reflect the actual surface shape of the solar-powered UAV to the greatest extent possible.

Description

Technical Field

[0001] This invention belongs to the field of aerodynamic optimization technology for unmanned aerial vehicles (UAVs), and particularly relates to a layered aerodynamic optimization method for micro solar-powered UAVs. Background Technology

[0002] Existing UAV aerodynamic optimization mainly falls into parametric optimization and non-parametric optimization. Parametric optimization is further divided into stochastic optimization and gradient optimization. Chinese invention patent application CN114154434A proposes a multi-constraint refined aerodynamic optimization design method that optimizes existing shapes based on gradient information through deformation. During the optimization process, the results are compared with the previous aerodynamic result to determine the correctness of the optimization gradient. Chinese invention patent CN110851912A discloses a fast optimization method based on FFD, which is also based on existing configurations. Currently, there is no systematic shape design method for micro fixed-wing UAVs, making it impossible to directly perform gradient-information-based deformation optimization. Furthermore, existing parametric description methods struggle to balance the number of optimization parameters and the shape envelope, potentially leading to local optima due to the limitations of the description methods themselves. Summary of the Invention

[0003] To address the aforementioned problems in existing technologies, this invention employs target parameters as a guide for aerodynamic optimization design of unmanned aerial vehicles (UAVs). Starting with desired performance indicators, it rapidly seeks the optimal target airfoil through hierarchical / layered optimization, and then performs nodal deformation of this airfoil based on gradient information. This aerodynamic optimization method does not rely on a fixed configuration, reducing the number of gradient deformation iterations and thus minimizing optimization overhead. Simultaneously, it maximizes the coverage of design conditions and reflects the actual surface shape of the solar-powered UAV to the greatest extent possible.

[0004] The technical solution of the present invention is as follows:

[0005] A hierarchical aerodynamic optimization method for a micro solar-powered unmanned aerial vehicle (UAV) includes the following steps:

[0006] S1: Perform two-dimensional airfoil parameterization description of the UAV;

[0007] The parametric description of the two-dimensional airfoil includes: dividing the two-dimensional airfoil into a thickness function and a mid-curvature function, which are drawn by spline curves. The constraints include leading edge radius, maximum thickness, maximum thickness position, maximum camber, maximum camber position, trailing edge angle, and trailing edge concave control.

[0008] S2: Based on the two-dimensional airfoil parameters obtained in step S1, perform multi-objective genetic algorithm optimization;

[0009] S3: Perform adjoint solution on the multi-objective genetic algorithm optimization results in step S2 to obtain optimization target gradient information, and perform freeform optimization of the two-dimensional airfoil based on the gradient information, and incorporate the results into the airfoil database;

[0010] S4: Based on the actual needs of the micro solar-powered UAV, a three-dimensional airfoil parameterization description is performed. The airfoil database is called as the selection factor to establish a three-dimensional airfoil parameterization model. The three-dimensional airfoil parameters include leading edge sweep angle, trailing edge sweep angle, twist angle, dihedral angle, span, and chord length.

[0011] S5: The Kriging model is used for surrogate model optimization. Global and local variables, constraints and objectives are set. A multi-objective genetic algorithm is used to solve the comprehensive trade-off design of maximizing the aerodynamic lift coefficient, maximizing the lift-to-drag ratio and minimizing the nose-down moment.

[0012] S6: For the optimization results of the surrogate model in step S5, perform adjoint solution to obtain the optimization target gradient information, and perform freeform optimization of the three-dimensional airfoil based on the gradient information.

[0013] Furthermore, step S2 specifically includes:

[0014] S21: Based on the two-dimensional airfoil parameters obtained in step S1, the initial sample is selected using the Latin hypercube experimental design;

[0015] S22: Perform two-dimensional airfoil population calculation;

[0016] S23: Sort the calculation results of step S22 by Pareto order;

[0017] S24: Use a multi-objective genetic algorithm to complete the population iteration of heredity, mutation, and crossover until the objective converges.

[0018] Furthermore, step S22 specifically includes:

[0019] Bezier spline curves are used to model the airfoil leading edge radius, maximum thickness location, maximum thickness, maximum camber location, maximum camber, trailing edge angle, and trailing edge concave. The resulting model is divided into finite volume sections, and aerodynamic calculations are performed using a turbulence model to solve for the target indices of airfoil lift coefficient, airfoil lift-to-drag ratio, and airfoil pitching moment.

[0020] Furthermore, step S5 specifically includes:

[0021] S51: Characterizing three-dimensional airfoil geometry through parametric description;

[0022] S52: Initial samples were selected using a Latin hypercube design experiment;

[0023] S53: Perform a three-dimensional high-confidence CFD solution for the initial sample;

[0024] S54: Substitute the solution results of step S53 into the co-kriging surrogate model framework to construct the initial function model, and complete the analysis and variance prediction of the function model by solving the maximum likelihood estimation function.

[0025] S55: Iteratively verify the variance of the maximum variance position in the whole period and the variance of the maximum variance position between local samples in the variance prediction results of step S54. The convergence condition of the iteration is based on the change of the predicted position of the function model. After the convergence condition is reached within the limit, step S6 is executed.

[0026] Furthermore, the limiting conditions are as follows: the maximum thickness position, maximum thickness, maximum curvature position, and maximum curvature parameters are limited based on the structural and material characteristics of the UAV to control the external shape boundary.

[0027] Furthermore, the convergence condition is as follows: when the variance no longer decreases further, the convergence condition is determined to be met, based on the comparison between the solution calculation results and the prediction results of the surrogate model framework, within the constraints of the UAV's geometric shape.

[0028] Furthermore, the three-dimensional high-confidence CFD solution in step S53 specifically includes:

[0029] S531: Secondary development is completed based on the script compilation method of modeling software to establish a three-dimensional airfoil parametric model;

[0030] S532: Perform high-confidence aerodynamic calculations on the three-dimensional airfoil parameterized model based on the finite volume method, including three-dimensional mesh generation and three-dimensional mesh calculation. The three-dimensional mesh generation is run by script, and the three-dimensional mesh calculation is solved.

[0031] Furthermore, the results of the three-dimensional mesh calculation in step S532 are: lift coefficient, drag coefficient, lift-to-drag ratio, and pitching moment coefficient.

[0032] Furthermore, step S6 specifically includes:

[0033] By optimizing the results of the surrogate model, we carry out free deformation optimization based on mesh deformation. We analyze the sensitivity of each mesh node through the adjoint solver, calculate the influence at each location, delineate the influence boundary, retain only the constraints in gradient optimization, and perform terminal optimization based on the constraints and gradient information to achieve the global optimal result.

[0034] Compared with existing technologies, the present invention has the following advantages:

[0035] This invention proposes a hierarchical aerodynamic optimization method for micro solar-powered unmanned aerial vehicles (UAVs). Based on two levels (three-dimensional and three-dimensional), it selects initial sample points using the Latin hypercube method, and combines calculation methods with different confidence levels, as well as using gradient information to determine the optimization direction. This greatly improves optimization efficiency, avoids reliance on the designer's experience, and can extend the optimization method's application domain by introducing constraints according to actual needs, thus efficiently completing accurate optimization. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the embodiments will be briefly described below. Referring to the accompanying drawings will provide a clearer understanding of the features and advantages of the present invention. The drawings are illustrative and should not be construed as limiting the present invention in any way. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort. Wherein:

[0037] Figure 1 This is a schematic flowchart of a layered aerodynamic optimization method for a micro solar-powered unmanned aerial vehicle according to the present invention;

[0038] Figure 2 The flowchart below shows the hierarchical optimization framework of the two-dimensional high-confidence multi-objective genetic algorithm used in this method.

[0039] Figure 3 The flowchart shows the three-dimensional surrogate model-based hierarchical optimization framework of this method. Detailed Implementation

[0040] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments of the present invention and the features thereof can be combined with each other.

[0041] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0042] This invention employs a hierarchical optimization method, which involves a two-step optimization process. The first step uses a genetic algorithm to select the optimal population and the optimal sample. The second step involves performing an adjoint solution on the current optimal sample to obtain gradient information and perform free deformation optimization based on mesh deformation. Furthermore, this invention employs a step-by-step optimization method, first optimizing the two-dimensional airfoil and then optimizing the three-dimensional airfoil according to specific practical requirements.

[0043] The optimization method is as follows:

[0044] Step 1: Select the genetic algorithm as the two-dimensional airfoil optimization method.

[0045] Step 2: Describe the two-dimensional airfoil shape by dividing the airfoil into a thickness function and a mid-curvature function, which are drawn by spline curves. The constraints include leading edge radius, maximum thickness, maximum thickness location, maximum camber, maximum camber location, trailing edge angle, and trailing edge inflection control.

[0046] Step 3: Divide the space into finite volumes based on the constructed airfoil, and use the turbulence model to perform aerodynamic calculations. Extract the airfoil lift coefficient, airfoil lift-to-drag ratio, and airfoil pitching moment from the solution results, sort them by Pareto order, and complete the population iteration of genetics, mutation, and crossover until the target converges. Specifically, the implementation is as follows: For initial sample selection, a Latin hypercube experimental design is used to address the issues of stratification, sampling, and randomized sampling. For aerodynamic calculations, a high-confidence finite volume method is employed. Computational fluid dynamics based on the finite volume method first requires spatial meshing of the computational domain and computational entities, with the meshing criteria tested through mesh independence verification. For target selection, key indicators such as lift coefficient, drag coefficient, and pitching moment coefficient are considered. The optimization method primarily employs a multi-objective genetic algorithm based on an elite retention strategy, using natural selection, superiority crossbreeding, and probabilistic mutation to iterate the population and design the envelope of the initial optimization solution. Regarding the optimization process, automated calculation and iteration are implemented. The meshing step utilizes scripts to automate the process from parametric modeling to mesh file generation. Simultaneously, the flow solution process is also completed using TUI (Automated User Interface), avoiding human intervention and reducing manual effort, iteration time, and more efficiently completing the optimization iteration.

[0047] Step 4: After completing the envelope of the initial optimization solution, it is necessary to perform adjoint optimization design based on gradient descent to complete the final optimization of non-optimal results caused by cognitive or representational deficiencies. After completing the two-dimensional airfoil optimization, the aerodynamic shape design optimization of the three-dimensional airfoil is performed according to the actual shape of the UAV, and the two-dimensional airfoil is used as the reference parameter to substitute into the three-dimensional airfoil for parameterization description.

[0048] Step 5: For the three-dimensional airfoil optimization problem, establish a parametric model of the shape, call the airfoil database as the selection factor, and design parameters including leading edge sweep angle, trailing edge sweep angle, twist angle, dihedral angle, span, and chord length. Use macro commands to complete the modeling.

[0049] Step 6: A surrogate model-based method is used to complete the three-dimensional airfoil optimization. The three-dimensional airfoil optimization problem presents a trade-off between computational cost and accuracy. High-confidence aerodynamic calculations require significant computational resources, while low-confidence calculations often result in large errors and easily miss the optimal point. To reconcile the issues of optimization accuracy and computational cost, this invention employs a co-kriging method. Optimization is achieved by utilizing samples with different confidence levels, setting global and local variables, constraints, and objectives, and using a multi-objective algorithm to obtain a comprehensive design that maximizes the aerodynamic lift coefficient, lift-to-drag ratio, and minimizes the nose-down moment. First, the geometry is characterized through parametric description. The Latin hypercube design experiment selected initial samples. For these initial samples, a high-confidence calculation method was chosen. Secondary development of the modeling software was achieved through script compilation using VBA and TCL environments. A descriptive method including leading-edge functions, chord length distribution functions, torsional distribution functions, and inverse functions was constructed to model parameters such as the 3D leading edge, chord length, torsion angle, and inverse angle. Automatic generation from parameters to the model was completed. Then, high-confidence aerodynamic calculations based on the finite volume method were performed, including 3D mesh generation and calculation. 3D mesh generation was also executed via scripts, and the flow solution process was completed using TUI, automatically optimizing and iterating by setting 3D calculation variables. After completing the calculations for the initial samples, the model was substituted into the surrogate model framework. An initial function model is constructed, and the model is analyzed and variance is predicted by solving the maximum likelihood estimation function. Iterative verification is performed on the parts with large variance. The calculation method of the iteration point is as follows: the variance verification of the maximum variance position of the whole cycle is carried out by high confidence aerodynamic calculation based on commercial fluid dynamics numerical simulation software. The calculation process still uses the automatic solution framework. For the maximum variance position between local samples, a low confidence fast solution method based on vortex lattice method is used. This method can quickly complete the local verification of the model without affecting the solution of the optimal point of the whole model. The iteration convergence condition is based on the change of the predicted position of the model. After the convergence condition is reached within the constraint range, step 7 is started.

[0050] Step 7: For the optimization results of the surrogate model, continue to carry out free deformation optimization based on mesh deformation. Analyze the sensitivity of each mesh node through the adjoint solver, calculate the influence at each position, and delineate the influence boundary. At this time, discard the standard parameterization constraints and perform gradient optimization, retaining only the constraints to avoid local optimization caused by fixed parameter description. Perform terminal optimization based on the constraints and gradient information to achieve the global optimal result.

[0051] The final results were verified through experimental design (see Table 1): For the two-dimensional low Reynolds number airfoil optimization problem, compared with the publicly disclosed airfoil NACA0015, the lift-to-drag ratio was improved by 57%, and the maximum lift coefficient was improved by 78%; for the three-dimensional low Reynolds number medium aspect ratio fixed wing shape optimization problem, the wing loading was reduced by 15%, and the lift-to-drag ratio was improved by 27%. This demonstrates that the optimization design method can effectively improve the aerodynamic performance of UAVs.

[0052] Table 1

[0053] Airfoil name Lift coefficient drag coefficient Rise-to-drag ratio Three-dimensional lift-to-drag ratio Remark NACA23012 0.7189 0.0234 30.7 8.12 Re=80000 Optimized airfoil 1.28 0.0265 48.2 10.31 Re=80000

[0054] The above description is merely an example of the implementation of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A hierarchical aerodynamic optimization method for a micro solar-powered unmanned aerial vehicle, characterized in that, Includes the following steps: S1: Perform a two-dimensional airfoil parameterization description on the UAV. The two-dimensional airfoil parameterization description includes: dividing the two-dimensional airfoil into a thickness function and a mid-curvature function, which are drawn by spline curves. The constraints include leading edge radius, maximum thickness, maximum thickness position, maximum camber, maximum camber position, trailing edge angle, and trailing edge concave control. S2: Based on the two-dimensional airfoil parameters obtained in step S1, perform multi-objective genetic algorithm optimization; S3: Perform adjoint solution on the multi-objective genetic algorithm optimization results in step S2 to obtain optimization target gradient information, and perform freeform optimization of the two-dimensional airfoil based on the gradient information, and incorporate the results into the airfoil database; S4: Based on the actual needs of the micro solar-powered UAV, a three-dimensional airfoil parameterization description is performed. The airfoil database is called as the selection factor to establish a three-dimensional airfoil parameterization model. The three-dimensional airfoil parameters include leading edge sweep angle, trailing edge sweep angle, twist angle, dihedral angle, span, and chord length. S5: The Kriging model is used for surrogate model optimization. Global and local variables, constraints and objectives are set. A multi-objective genetic algorithm is used to solve the comprehensive trade-off design of maximizing the aerodynamic lift coefficient, maximizing the lift-to-drag ratio and minimizing the nose-down moment. S6: For the optimization results of the surrogate model in step S5, perform adjoint solution to obtain the optimization target gradient information, and perform freeform optimization of the three-dimensional airfoil based on the gradient information. Step S2 specifically includes: S21: Based on the two-dimensional airfoil parameters obtained in step S1, the initial sample is selected using the Latin hypercube experimental design; S22: Perform two-dimensional airfoil population calculation; S23: Sort the calculation results of step S22 by Pareto order; S24: Use a multi-objective genetic algorithm to complete the population iteration of heredity, mutation, and crossover until the objective converges; Step S22 specifically includes: Bezier spline curves are used to model the airfoil leading edge radius, maximum thickness location, maximum thickness, maximum camber location, maximum camber, trailing edge angle, and trailing edge concave. The resulting model is divided into finite volume sections, and aerodynamic calculations are performed using a turbulence model to solve for the target indices of airfoil lift coefficient, airfoil lift-to-drag ratio, and airfoil pitching moment.

2. The layered aerodynamic optimization method for micro solar-powered unmanned aerial vehicles according to claim 1, characterized in that, Step S5 specifically includes: S51: Characterizing three-dimensional airfoil geometry through parametric description; S52: Initial samples were selected using a Latin hypercube design experiment; S53: Perform a three-dimensional high-confidence CFD solution for the initial sample; S54: Substitute the solution result of step S53 into the surrogate model framework to construct the initial function model, and complete the analysis and variance prediction of the function model by solving the maximum likelihood estimation function. S55: Iteratively verify the variance of the maximum variance position in the whole period and the variance of the maximum variance position between local samples in the variance prediction results of step S54. The convergence condition of the iteration is based on the change of the predicted position of the function model. After the convergence condition is reached within the limit, step S6 is executed.

3. The layered aerodynamic optimization method for micro solar-powered unmanned aerial vehicles according to claim 2, characterized in that, The constraints are as follows: the maximum thickness location, maximum thickness, maximum curvature location, and maximum curvature parameters are restricted based on the structural and material characteristics of the UAV to control the external shape boundary.

4. The layered aerodynamic optimization method for micro solar-powered unmanned aerial vehicles according to claim 3, characterized in that, The convergence condition is as follows: when the variance no longer decreases further, the convergence condition is determined to be met, based on the comparison between the solution calculation results and the prediction results of the surrogate model framework, within the constraints of the UAV's geometric shape.

5. The layered aerodynamic optimization method for micro solar-powered unmanned aerial vehicles according to claim 2, characterized in that, The three-dimensional high-confidence CFD solution in step S53 specifically includes: S531: Secondary development is completed based on the script compilation method of modeling software to establish a three-dimensional airfoil parametric model; S532: Perform high-confidence aerodynamic calculations on the three-dimensional airfoil parameterized model based on the finite volume method, including three-dimensional mesh generation and three-dimensional mesh calculation. The three-dimensional mesh generation is run by script, and the three-dimensional mesh calculation is solved.

6. The layered aerodynamic optimization method for micro solar-powered unmanned aerial vehicles according to claim 5, characterized in that, The results of the three-dimensional mesh calculation in step S532 are: lift coefficient, drag coefficient, lift-to-drag ratio, and pitching moment coefficient.

7. The layered aerodynamic optimization method for micro solar-powered unmanned aerial vehicles according to claim 1, characterized in that, Step S6 specifically includes: By optimizing the results of the surrogate model, we carry out free deformation optimization based on mesh deformation. We analyze the sensitivity of each mesh node through the adjoint solver, calculate the influence at each location, delineate the influence boundary, retain only the constraints in gradient optimization, and perform terminal optimization based on the constraints and gradient information to achieve the global optimal result.

Images