A ducted tail-sitter unmanned aerial vehicle aerodynamic optimization design method
By employing multi-condition differentiated parallel flow field simulation and hierarchical multi-objective genetic algorithm optimization, the problem of the three-mode aerodynamic performance differences of ducted tail-mounted UAVs was solved, achieving performance balance and stability under all flight conditions, and improving vertical take-off and landing and cruise efficiency.
Patent Information
- Application Number
- CN202611544795.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-08
- Publication Date
- 2026-11-03
AI Technical Summary
Existing technologies struggle to simultaneously achieve the three-mode aerodynamic performance of ducted tail-mounted UAVs, especially given the significant differences in aerodynamic characteristics during vertical takeoff and landing, attitude transition, and horizontal cruise phases. Furthermore, the aerodynamic coupling between the duct, propeller, and built-in tail rudder is severe, resulting in poor flight stability and high computational costs.
Multi-condition differentiated parallel flow field simulation and hierarchical multi-objective genetic algorithm optimization are adopted to configure independent boundary conditions for vertical take-off and landing, attitude transition and horizontal cruise modes respectively. The aerodynamic response is quickly predicted by the surrogate model. The torsional moment of vertical take-off and landing and the pitch moment of attitude transition are set as the main optimization objectives, and the lift-to-drag ratio of horizontal cruise is set as the secondary optimization objective. Automatic iterative optimization is performed to output the optimal unmanned structure parameters.
It achieves balanced performance across all flight conditions, ensuring takeoff and landing stability and cruise efficiency. The vertical takeoff and landing torsional moment and attitude transition pitch moment meet design requirements, and the horizontal cruise lift-to-drag ratio is improved by 4.35%, extending endurance.
Smart Images
Figure CN123031637A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of aerodynamic shape optimization design, specifically relating to an aerodynamic optimization design method for a ducted tail-mounted unmanned aerial vehicle (UAV). Background Technology
[0002] Ductless tail-seat UAVs are vertical takeoff and landing (VTOL) aircraft that switch between vertical takeoff and landing and horizontal cruise modes by changing the fuselage attitude. Compared to tiltrotor and compound-wing UAVs, they eliminate complex tilting mechanisms and redundant power systems, resulting in simpler structures, lighter weight, and lower cost. Furthermore, the use of ducted propellers solves the problems of poor safety, high noise, and weak disturbance resistance associated with traditional exposed rotors, making them highly valuable for engineering applications in small to medium-sized high-speed VTOL UAVs.
[0003] However, the aerodynamic design of ducted tail-seat UAVs faces a core contradiction that existing methods have failed to resolve: this type of configuration exhibits three flight modes with vastly different aerodynamic characteristics—vertical takeoff and landing, attitude transition, and horizontal cruise—and suffers from severe aerodynamic coupling among the duct, propeller, and internal tail rudder, resulting in complex multi-component interference effects. Specifically: First, if the configuration prioritizes cruise lift-to-drag ratio optimization, it can lead to excessive propeller-induced torsional moment and excessive control surface load during vertical takeoff and landing, threatening takeoff and landing safety; conversely, prioritizing control torque constraints during takeoff and landing significantly deteriorates cruise aerodynamic efficiency. Second, there is strong mutual interference between the propeller downwash, the duct lip flow, and the control surface deflection flow field. Existing empirical configurations struggle to quantify and mitigate the additional drag and torque disturbances caused by coupling, resulting in localized airflow separation and poor flight stability. Third, the attitude transition phase (the leap from vertical takeoff and landing to horizontal cruise) has long been neglected by existing design methods as an independent dynamic mode, and the pitch moment during this transition phase lacks quantitative control mechanisms.
[0004] To address the aforementioned multimodal aerodynamic contradictions, existing optimization schemes mainly fall into two categories. The first is single-condition oriented optimization: this focuses on optimal aerodynamic design for only one of the operating conditions—vertical takeoff and landing or horizontal cruise—essentially sacrificing performance in one mode to improve performance in another, failing to cover the entire flight profile. The second is multi-objective weighted optimization: this linearly combines multiple objectives such as torsional moment, pitch moment, and lift-to-drag ratio into a single comprehensive function by setting weight coefficients, then uses intelligent algorithms such as genetic algorithms to find the optimal solution. However, this method has inherent technical flaws: torsional moment, pitch moment, and lift-to-drag ratio have different dimensions and orders of magnitude, and the setting of weight coefficients lacks objective physical basis. This easily leads to the algorithm converging to a mediocre solution where all indicators compromise and none are met, especially failing to guarantee rigid safety constraints on attitude moments during takeoff, landing, and transition phases. The optimization results often fail in actual flight due to uncontrolled moments. Furthermore, regardless of which approach is adopted, if high-precision CFD simulation is directly used as the fitness evaluation method for optimization iteration, the computational cost is completely unacceptable in engineering because each generation of individuals needs to complete the solution of the three-mode flow field, and a single optimization requires hundreds or thousands of simulations.
[0005] In summary, the existing technology lacks an aerodynamic optimization design method that can simultaneously take into account the three-modal aerodynamic performance of ducted tail-mounted UAVs, apply hard priority guarantees to attitude stabilization torques, and integrate surrogate model reliability constraints into the optimization process. Summary of the Invention
[0006] To address the aforementioned issues, this invention discloses a highly targeted, comprehensive, and adaptable aerodynamic optimization design method for ducted tail-mounted unmanned aerial vehicles (UAVs) that is compatible with three operating conditions. This method also considers the three-modal aerodynamic performance of ducted tail-mounted UAVs, while offering low computational cost and high computational accuracy.
[0007] To achieve the above objectives, the present invention adopts the following technical solution:
[0008] An aerodynamic optimization design method for a ducted tail-mounted unmanned aerial vehicle (UAV) includes the following steps:
[0009] S1, Initial Configuration Parametric Definition: For the initial shape of the ducted tail-mounted UAV, minor geometric parameters are removed, and a set of geometric design variables that play a decisive role in the duct aerodynamic efficiency, propeller-induced flow field, and tail rudder control torque are extracted. These geometric design variables include: duct lip diameter D. in duct outlet diameter D out The geometric design variables include: duct length L, duct cross-sectional thickness t, propeller axial position l1, duct tail rudder axial position l2, and duct tail rudder chord length c. Based on these geometric design variables, a parametric modeling system is established, and the value range of each geometric design variable is defined to form a design space.
[0010] S2, Multi-condition differentiated parallel flow field simulation: Based on the geometric design variables determined in step 1, complete the three-dimensional configuration modeling and calculate the mesh deformation. Use computational fluid dynamics simulation tools to independently configure boundary conditions, thrust loads and control surface deflection angles for the vertical take-off and landing mode, attitude transition mode and horizontal cruise mode respectively. Obtain the six-component data of the whole aircraft under each mode through three-mode parallel simulation, including the torsional moment of the vertical take-off and landing mode, the pitch moment of the attitude transition mode and the lift and drag of the horizontal cruise mode;
[0011] S3, Hierarchical Multi-Objective Iterative Optimization: Based on the simulation data obtained in step 2, a surrogate model is sampled and established within the design space, with the geometric design variables as input and the aerodynamic response of each mode as output. The optimization objectives are divided into primary and secondary levels, wherein the improvement of the torsional moment of the vertical take-off and landing mode and the pitch moment of the attitude transition mode are set as the primary optimization objectives, and the improvement of the lift-to-drag ratio of the horizontal cruise mode is set as the secondary optimization objectives. The primary optimization objectives have higher priority than the secondary optimization objectives. With the geometric design variables as optimization variables and the values of the constraint optimization variables not exceeding the range of the design space, a genetic algorithm is used to automatically iterate and optimize within the design space, outputting a set of geometric design variables that satisfy the constraints of the primary optimization objectives and make the secondary optimization objectives optimal, as the optimal unmanned aircraft configuration parameters.
[0012] Furthermore, in step 1, the value range of the geometric design variables is: 0.65m <D in <0.68m, 0.6m <D out <0.63m, 0.24m <L<0.26m、0.03m<t<0.04m、0.04m< l1<0.06m、0.1m< l2<0.12m、0.14m< c<0.16m。
[0013] Furthermore, in step 2, the boundary conditions for the three flight modes are configured as follows:
[0014] Vertical takeoff and landing mode: incoming flow velocity 0 m / s, total thrust of propeller momentum source 25 kgf, overall angle of attack 0°, all tail rudder surfaces uniformly deflected 5° to counteract the torsional torque generated by propeller counter-torque;
[0015] Attitude transition mode: incoming flow velocity 10 m / s, total propeller momentum source thrust 25 kgf, overall angle of attack 2°, elevator surfaces additionally deflected downward 5° to provide pitch moment for attitude transition;
[0016] Horizontal cruise mode: incoming flow velocity 42 m / s, total thrust of propeller momentum source 6 kgf, overall angle of attack 2°.
[0017] Furthermore, in step 2, the aerodynamic response data includes six components of the entire aircraft data under each mode: lift, drag, side force, roll moment, pitch moment, and yaw moment; wherein, the torsional moment in the roll direction is extracted in the vertical take-off and landing mode, the pitch moment is extracted in the attitude transition mode, and the lift and drag are extracted in the horizontal cruise mode for calculating the lift-to-drag ratio.
[0018] Furthermore, in step 2, the turbulence model uniformly adopts the k-ω SST two-equation model, and the spatial discretization adopts the second-order upwind scheme; the computational grid adopts a hybrid grid, with hexahedral grids in the boundary layer region and tetrahedral grids in the far-field region. The far-field boundary is 20 times the span from the aircraft surface, and the boundary layer is divided into 30 layers. The height of the first layer of the grid satisfies the wall y + < 1.
[0019] Furthermore, in step 2, the propeller defines a momentum source term in a separately divided volume grid region within the computational domain to simulate the momentum injection of the propeller disk into the airflow.
[0020] Furthermore, in step 3, the proxy model is established as follows: sample points are extracted in the design space using the Latin hypercube sampling method, and the parallel flow field simulation is performed on each sample point to obtain the corresponding aerodynamic response data, forming a training sample pair. Then, a neural network is used to fit the mapping relationship between the geometric design variables and the aerodynamic response data.
[0021] Furthermore, in step 3, the quantitative constraints of the primary optimization objective are: the torsional moment of the vertical take-off and landing mode ≥ 8.0 N·m, and the pitching moment of the attitude transition mode ≥ 5.0 N·m; the secondary optimization objective is: the lift-to-drag ratio of the entire aircraft in the horizontal cruise mode is improved by ≥ 4.0% compared with the initial shape.
[0022] Furthermore, in step 3, when comparing the merits of two individuals, it is first determined whether the primary optimization objective has met the quantitative constraints. Individuals that have not met the primary optimization objective are eliminated. Only when both individuals have met the primary optimization objectives are the values of their secondary optimization objectives compared, and the one with the better secondary optimization objective is retained.
[0023] Furthermore, in step 3, after outputting the optimal configuration parameters, a verification step is also included: the optimal geometric design variables output by optimization are substituted back into the parallel flow field simulation process of step 2 to independently verify the three-mode aerodynamic response; if the deviation between the verification result and the predicted value of the surrogate model exceeds the preset threshold, the sampling step is returned to supplement the training samples and the surrogate model is retrained, and iterative optimization is performed again until the verification is passed.
[0024] Beneficial effects:
[0025] (1) Compared with existing single-condition optimization methods, this invention uses three-modal parallel simulation to calculate the aerodynamic response of three conditions—vertical takeoff and landing, attitude transition, and horizontal cruise—on the same parameterized model, thus ensuring that the optimization process takes into account the requirements of the entire flight profile. Optimizing the cruise lift-to-drag ratio alone can lead to insufficient control torque during takeoff, landing, and transition phases; prioritizing takeoff and landing stability can significantly sacrifice cruise efficiency. This invention avoids this trade-off and achieves balanced performance across all flight conditions.
[0026] (2) This invention sets the torsional moment during the vertical takeoff and landing phase and the pitching moment during the attitude transition phase as the primary optimization targets, prioritizing indicators such as lift-to-drag ratio, to avoid attitude stability degradation due to the pursuit of efficiency. After optimization, the torsional moment of the control surfaces during takeoff and landing can reach more than 8.0 N·m, and the pitching moment during the attitude transition phase can reach more than 5.0 N·m, effectively suppressing takeoff and landing jitter and attitude abrupt changes during mode switching, ensuring attitude controllability and safety throughout the entire flight process.
[0027] (3) Under the premise of meeting the above attitude stability requirements, this invention optimizes the cruise lift-to-drag ratio. Actual test data shows that the optimized horizontal cruise lift-to-drag ratio increases from 5.80 in the initial configuration to 6.05, an increase of 4.35%, reaching the preset optimization target of ≥4%. The improved lift-to-drag ratio reduces cruise drag and power consumption, extends endurance and operating radius, and is beneficial for performing long-endurance, long-distance missions. Attached Figure Description
[0028] Figure 1 This is a flowchart illustrating the aerodynamic optimization process for a ducted tail-mounted UAV according to an embodiment of the present invention.
[0029] Figure 2 This is the initial aerodynamic shape of the tail-mounted ducted unmanned aerial vehicle according to an embodiment of the present invention;
[0030] Figure 3 Design variable D for the aerodynamic optimization design of the tail-mounted ducted unmanned aerial vehicle in this embodiment of the invention. in D out L;
[0031] Figure 4 The design variables l1, l2, t, and c are for the aerodynamic optimization design of the tail-mounted ducted unmanned aerial vehicle in this embodiment of the invention.
[0032] Figure 5 This is the curve showing the change of the objective function of this invention with the number of iterations in an embodiment of the invention.
[0033] Reference numerals: 1. Fuselage 2. Wing 3. Propeller 4. Duct 5. Control surface Detailed Implementation
[0034] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.
[0035] This embodiment discloses an aerodynamic optimization design method for a ducted tail-mounted unmanned aerial vehicle (UAV). Existing technologies mainly have three problems: First, the requirements for aerodynamic performance in the three flight conditions of vertical take-off and landing, attitude transition, and horizontal cruise are mutually restrictive. Traditional single-condition optimization can only take into account one of them and cannot take into account the overall performance. Second, the duct, propeller, and built-in tail rudder are close to each other and strongly coupled, resulting in complex airflow interference effects. It is difficult to accurately assess and mitigate these adverse effects by relying on human experience. Third, configuration parameters are usually manually trial-and-error and iteratively adjusted, which is inefficient and lacks intelligent optimization methods that automatically search for the optimal parameter combination under multiple constraints.
[0036] To address the aforementioned issues, this embodiment divides the aerodynamic optimization process into three steps: parameter definition, flow field simulation, and automatic optimization. First, initial configuration parameterization is performed. A set of independent and controllable design variables is extracted, focusing on key structural dimensions affecting duct efficiency, propeller flow field, and tail rudder control torque, to establish a unified parameterized model, providing stable input for subsequent simulations and optimization. Then, multi-condition differentiated parallel simulations are conducted. Based on the geometric parameters confirmed in the previous steps, 3D modeling and mesh deformation are completed. Boundary conditions, thrust loads, and control surface deflections are independently set for three modes: vertical takeoff and landing, attitude transition, and horizontal cruise. Flow field, aerodynamic forces, and control torque data for the entire flight profile are obtained through parallel computation, thereby revealing the influence of each parameter on the aerodynamic characteristics under different operating conditions. Finally, a hierarchical multi-objective genetic algorithm is used for iterative optimization. A surrogate model is built on the simulation data to replace the computational fluid dynamics simulation. The takeoff and landing torsional torque and transition pitch torque, which are related to attitude stability, are set as the primary optimization objectives, and the improvement of the cruise lift-to-drag ratio is set as the secondary optimization objective. The range of values for each design variable is limited. Then, the genetic algorithm with set population size, crossover and mutation probability and convergence criterion is used for automatic iterative looping. Under the premise of meeting the attitude torque requirements, the globally optimal configuration combination is searched. Finally, the optimal unmanned structure parameters that take into account the stability of the three working conditions and the cruise efficiency are output.
[0037] Specifically,
[0038] like Figure 2The figure shows the initial three-dimensional shape of the ducted tail-mounted UAV involved in this embodiment. The UAV adopts a tail-mounted layout, with the fuselage 1 serving as the main load-bearing structure and connecting all core components. In aerodynamic optimization, the cross-sectional shape and aspect ratio of the fuselage affect the overall drag and cruise lift-to-drag ratio of the aircraft. The wing 2 is a fixed wing surface arranged vertically on the fuselage. It provides the main lift in horizontal cruise mode and serves as a lateral support surface during vertical takeoff and landing. Its installation angle and aspect ratio are key optimization variables. The propeller 3 is located inside the duct and is the core power source providing thrust for vertical takeoff and landing and thrust for horizontal cruise. In the simulation, it is injected as a momentum source term, and its induced flow field is strongly coupled with the duct. The duct 4 is an annular shell structure surrounding the propeller. It is used to improve the propeller slipflow characteristics and increase hovering efficiency. Its lip shape, inner diameter, and axial position are the core design variables extracted in step 1. The control surfaces 5 are control surfaces built into the rear of the duct (usually in a cross-shaped or X-shaped layout). It is used to counteract the reverse torque (torsional moment) during vertical takeoff and landing, and to provide pitch moment during attitude transition. Its deflection angle is the key basis for the simulation configuration in step 2 and the main optimization objective in step 3.
[0039] like Figure 1 As shown, an aerodynamic optimization design method for a ducted tail-mounted unmanned aerial vehicle (UAV) includes the following steps:
[0040] Step 1, Define key geometric parameters: Extract a set of independent and controllable core geometric design variables from the structural dimensions related to duct aerodynamic efficiency, propeller-induced flow field, and tail rudder control torque, and establish a parametric modeling system. For example... Figures 3-4 As shown, the core geometric design variables include: duct lip diameter D in duct outlet diameter D out The following parameters are given: duct length L, duct cross-sectional thickness t, propeller axial position l1, duct tail rudder axial position l2, duct tail rudder chord length c.
[0041] Among them, the diameter D of the duct lip in Let D be the inner diameter of the duct inlet. Optimizing this variable can improve the smoothness of airflow into the duct, directly affecting the intake efficiency and anti-torque characteristics within the duct during hovering. The duct outlet diameter D is also considered. out The inner diameter of the duct outlet end is D. inThe expansion / contraction ratio of the duct is determined by several factors, affecting the slipstream acceleration and thrust during cruise. The duct length L is the total axial length of the duct along the X-axis; changes in length alter the frictional contact area between the duct inner wall and the propeller slipstream, affecting overall drag and thrust utilization. The duct cross-sectional thickness t is the thickness of the duct wall, affecting the duct's structural weight and the frictional drag of the airflow on the outer surface. The propeller axial position l1 is the propeller's installation position along the X-axis inside the duct, determining the relative distance between the propeller and the duct lip / outlet, crucial for suppressing flow separation and maximizing momentum injection. The tail rudder axial position l2 is the installation position of the tail control surfaces along the X-axis inside the duct; changes in position directly affect the control surface's handling efficiency in the slipstream (i.e., the magnitude of the generated torsional and pitching moments). The tail rudder chord c is the chord length of the tail control surfaces; the chord length determines the effective control area of the surfaces and, together with the deflection angle, determines the upper limit of the attitude control torque.
[0042] In traditional manual trial-and-error methods, UAVs have numerous configuration parameters (such as wing surface details, fuselage curves, and propeller twist distribution). Optimizing all of these would lead to an explosion in computational complexity and a chaotic optimization direction. Therefore, this embodiment focuses on the core parameters that dominate aerodynamic performance, eliminating secondary parameters.
[0043] Step 2, Multi-condition differentiated parallel flow field simulation: Based on the geometric parameters determined in Step 1, complete the three-dimensional configuration modeling and computational mesh generation. Using computational fluid dynamics simulation tools, independently configure boundary conditions for vertical take-off and landing mode, attitude transition mode, and horizontal cruise mode. Obtain flow field data, aerodynamic coefficients, and control torques of the entire flight profile through parallel simulation.
[0044] In this embodiment, a hybrid mesh is used, with hexahedral meshes for the boundary layer and tetrahedral meshes for the far field, totaling approximately 6 million meshes. The height of the first layer mesh is set to ensure y + <1.0, with approximately 30 boundary layers. It should be noted that a separate volumetric mesh region is allocated for the propeller location to define the momentum source. That is, the propeller is not modeled as a rotating solid, but rather momentum is injected as a momentum source term to simulate the effect of the propeller disk on the airflow.
[0045] Because the three flight modes have different physical characteristics, they are treated separately. In this embodiment, the boundary conditions for the three flight modes are as follows:
[0046] Vertical Takeoff and Landing (VTOL) Mode: Incoming flow velocity 0 m / s, simulating pure hovering or low-speed climb without forward flight speed; total propeller thrust 25 kgf; overall angle of attack 0°; all control surfaces (typically cross-shaped or X-shaped in tail-seat configurations; this embodiment uses a star-shaped control surface, applicable to other configurations) are uniformly deflected 5° clockwise (viewed from rear to front). It should be noted that during VTOL, propeller rotation generates a counter-torque; without control, the fuselage will rotate around its longitudinal axis. The 5° clockwise control surface deflection generates a lateral (torsional) moment to counteract this counter-torque, maintaining the fuselage in a stable rolling attitude. It should be noted that this torsional moment value corresponds to ≥8.0 N·m in the main optimization objective of step 3.
[0047] Attitude transition mode: Incoming flow velocity 10 m / s, simulating an aircraft that has already achieved a certain forward speed during the climb transition; thrust remains at 25 kgf; overall angle of attack 2° (small angle of attack acceleration state); the two elevator surfaces in the horizontal direction are additionally deflected downwards (i.e., in the negative direction of the Z-axis of the body coordinate system) by 5°. Tail-seat UAVs typically use a climb to complete the mode transition, at which point the airflow has begun to wash over the wings and tail, and the control surfaces begin to have an effect. The additional 5° downward deflection is to provide a forward pitch moment, allowing the aircraft to smoothly transition from a small angle of attack to a horizontal state, avoiding sudden changes in pitch attitude that could lead to loss of control.
[0048] Horizontal cruise mode: Incoming flow velocity 42 m / s; thrust drops to 6 kgf (cruise only needs to overcome drag, lift is mainly provided by the fuselage / wing); overall angle of attack 2°. At this time, the ducted propeller is under low load, its main task being to reduce drag and maintain thrust. It should be noted that 42 m / s is the design cruise speed for this aircraft model; the lift-to-drag ratio extracted under this condition corresponds to the sub-optimization objective in step 3.
[0049] During aerodynamic simulation, the k-ωSST two-equation model is uniformly used for turbulence modeling, and a second-order upwind scheme is adopted for spatial discretization. Six-component data of the entire aircraft under three different flight modes are extracted as standard outputs and passed to the surrogate model. After simulation convergence, the standard output is the six-component data of the entire aircraft: lift (normal force), drag (axial force), lateral force (lateral force), roll moment, pitch moment, and yaw moment. Lift (normal force), drag (axial force), and lateral force (lateral force) are the components of the aerodynamic resultant force on the vertical, longitudinal, and transverse axes of the airflow coordinate system, respectively; roll moment, pitch moment, and yaw moment are the moment components of the aerodynamic resultant force about a reference point (usually the UAV's center of gravity) on the body axis coordinate system. Lift and drag are calculated by projecting the forces of the body axis system onto the wind axis system. The actual output of the surrogate model should be the normal force, axial force, and lateral force; lift and drag are obtained through angle-of-attack conversion, thus yielding the lift-to-drag ratio. In this embodiment, the VTOL mode outputs the torsional moment in the roll direction, the attitude transition mode outputs the pitch moment, and the cruise mode outputs lift and drag. Side force and yaw moment are not used as optimization targets, but only for flow field symmetry and numerical error verification. These data are input as real labels to the surrogate model in step 3 to train and fit the mapping relationship between the design variables and the aerodynamic response. In this embodiment, there is a nonlinear mapping relationship and coupling effect between the seven geometric design variables and the six-component aerodynamic response, which cannot be described by explicit algebraic expressions. During the aerodynamic simulation stage, the six-component data (lift, drag, side force, roll moment, pitch moment, and yaw moment) are obtained by integrating the pressure and viscous shear stress of the entire aircraft surface mesh. Specifically, the VTOL mode outputs the torsional moment in the roll direction, the attitude transition mode outputs the pitch moment, and the cruise mode outputs lift and drag; while the side force and yaw moment are not used as optimization targets, but only for flow field symmetry and numerical error verification. During the surrogate model construction phase, the six-component data obtained from the above simulation are used as the real labels to input into the surrogate model for training and fitting the mapping relationship between the design variables and the aerodynamic response; in subsequent optimization iterations, the surrogate model directly predicts the six-component data under the corresponding geometric parameters.
[0050] Step 3, hierarchical multi-objective genetic algorithm parameter iterative optimization: Based on the data obtained in Step 2, a surrogate model is established to replace the computational fluid dynamics simulation tool. The primary optimization objectives are to increase the torsional moment of the vertical takeoff and landing mode and the pitch moment of the attitude transition mode, while the secondary optimization objective is to increase the lift-to-drag ratio of the horizontal cruise mode. The value range of the core geometric design variables is limited. A genetic algorithm with defined population size, crossover probability, mutation probability, and iterative convergence criteria is used for automatic iterative optimization. Under the premise of satisfying the attitude moment improvement, the globally optimal geometric configuration combination is searched, and the optimal unmanned vehicle configuration parameters that balance stability under the three operating conditions and cruise aerodynamic efficiency are output. Specifically, this includes:
[0051] Step 3.1, establish the proxy model.
[0052] In addressing the issues that existing technologies typically rely on manual trial and error and empirical adjustments to determine configuration parameters, requiring remodeling for each adjustment and involving numerous design variables that make it difficult for humans to accurately assess the true contribution of each variable to performance, this embodiment introduces a surrogate model. This model requires only a small number of samples to build and can replace CFD for rapid prediction of aerodynamic performance.
[0053] First, sample data is acquired by randomly sampling in the design space using Latin hypercube sampling to obtain uniformly distributed initial sample points. Each sample point corresponds to a complete set of 7 geometric parameters (D). in D out (L, t, l1, l2, c). For each sample point, perform 3D modeling, mesh deformation, and three-modal parallel CFD simulation in step S2 to obtain the full-aircraft six-component data (lift, drag, side force, roll moment, pitch moment, yaw moment) for the three flight modes under this combination, forming parameter combination-aerodynamic response sample pairs, which serve as training data for the surrogate model. Using seven geometric parameters as input and the six-component data as output, train a neural network to fit the mapping relationship between the parameters and aerodynamic characteristics. After training, the surrogate model is directly used for rapid prediction in subsequent optimization processes.
[0054] Step 3.2: Iterative optimization using a hierarchical multi-objective genetic algorithm.
[0055] Based on the surrogate model, a hierarchical multi-objective genetic algorithm is used to search for the optimal combination of geometric parameters, including:
[0056] Step 3.21, Determine primary and secondary optimization objectives. Since different physical quantities have different dimensions, and there is a lack of basis for weighting, this embodiment adopts a hierarchical approach, dividing them into primary and secondary optimization objectives:
[0057] The primary optimization objective, i.e., the hard constraint that must be achieved:
[0058] ① Increase the torsional torque of the vertical take-off and landing mode, with a desired torsional torque ≥ 8.0 N·m (related to the vertical take-off and landing mode, 8.0 N·m comes from the counter torque brought to the whole aircraft by the motor driving the propeller during the take-off of the UAV), so as to ensure the stability of the roll attitude of the take-off and landing mode.
[0059] ② Increase the pitch moment of the attitude transition mode, with a desired pitch moment ≥ 5.0 N·m (related to the attitude transition mode; when the pitch moment is 5.0 N·m, the time limit of the attitude transition process needs to be considered, and the corresponding angular acceleration needs to be reached) to ensure the stability of the attitude during mode switching.
[0060] The secondary optimization objective, after satisfying the primary objective, seeks a better performance indicator: improving the overall lift-to-drag ratio in cruise mode, with a desired improvement of 4.0% compared to the initial shape. The lift-to-drag ratio is calculated as follows: cruise power is obtained from the UAV's endurance and battery capacity; flight drag is determined from flight power; since lift equals gravity, improving the lift-to-drag ratio means reducing flight drag, reducing cruise power, and increasing endurance. The 4.0% improvement is determined by increasing endurance.
[0061] The constraint variables are 7 geometric parameters, and the range of variable values does not exceed the range of values in the sample space.
[0062] When comparing the merits of two individuals, the rule of prioritizing the primary objective is followed: only when both primary objectives are met can the increase in lift-to-drag ratio be compared; individuals that fail to meet their primary objectives will be eliminated regardless of their lift-to-drag ratio.
[0063] Step 3.22, set the genetic algorithm parameters: The main genetic algorithm parameters are defined as follows: population size 110, crossover probability 0.78, mutation probability 0.12, and the iteration convergence threshold is that the fluctuation of the objective function is <0.3% for 15 consecutive generations.
[0064] Step 3.23, execute the iteration and output the optimal parameters: The iterative process of the genetic algorithm is as follows: First, randomly generate 110 sets of initial parameter individuals within the design space, and call the surrogate model to calculate the aerodynamic response of each individual in three modes; then, sort and select according to the above hierarchical rules, perform crossover and mutation on the selected individuals, and generate a new generation of population. Repeat the above process until the target fluctuation is less than 0.3% for 15 consecutive generations, and finally output the geometric parameter combination of the optimal individual.
[0065] like Figure 5 The figure shows the iterative trajectory of the genetic algorithm. The genetic algorithm converged at approximately 21 generations. During the iteration process, the torsional moment of the vertical takeoff and landing mode met the hard constraint of ≥8.0 N·m throughout the entire process, and the pitch moment of the attitude transition mode significantly improved from the initial approximately 5.3 N·m and stabilized, verifying the priority achievement of the primary optimization objective (attitude stability). On this basis, the lift-to-drag ratio of the horizontal cruise mode steadily improved from the initial 5.80 to the converged value of 6.05, an improvement of 4.35%, successfully meeting the secondary optimization index (cruise efficiency) of ≥4.0%. This convergence curve intuitively demonstrates that the hierarchical multi-objective optimization method of this application can synergistically take into account the aerodynamic performance of each mode.
[0066] In summary, existing methods typically perform single-point optimization only for steady-state cruise or hovering. Even when performing multi-state simulations, they often combine forces and moments from each state into a single target definition. This application addresses the three flight stages unique to ducted tail-seat UAVs, assigning clearly defined physical boundary conditions and target extraction methods for each. The roll torque is extracted through the vertical takeoff and landing mode, the pitch torque through the attitude transition mode, and the lift-to-drag ratio through the horizontal cruise mode. The attitude transition mode is often neglected or simplified to an intermediate interpolation state in most ducted tail-seat UAV aerodynamic optimization literature. This application elevates the transition mode to an independent optimization object alongside hovering and cruise, providing specific boundary condition configurations and quantitative constraints on pitch torque, filling the gap in existing technologies regarding the lack of control over pitch torque during the transition phase. Furthermore, this application does not employ general UAV full-parameter optimization (wing aspect ratio, fuselage slenderness ratio, airfoil camber, etc.), but instead selects seven aggregate variables based on the specific configuration of the ducted tail-seat UAV. In particular, by incorporating propeller position and rudder position (l1, l2) into the optimization variables, the interference efficiency between the slipstream and the rudder surface is directly affected, which is more accurate than the traditional method of only optimizing the duct diameter and length.
[0067] The embodiments described above are merely illustrative of the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solutions based on the technical concept proposed in this invention shall fall within the scope of protection of this invention.
Claims
1. An aerodynamic optimization design method for a ducted tail-mounted unmanned aerial vehicle (UAV), characterized in that, Includes the following steps: Step 1: For the initial shape of the ducted tail-mounted UAV, extract the geometric design variables, including: duct lip diameter D. in duct outlet diameter D out Duct length L, duct cross-sectional thickness t, propeller axial position l1, duct tail rudder axial position l2, duct tail rudder chord length c; Step 2: Based on the geometric design variables, perform three-dimensional configuration modeling and calculate mesh deformation. Independently configure boundary conditions for vertical take-off and landing mode, attitude transition mode, and horizontal cruise mode respectively. Obtain the aerodynamic response data of the whole aircraft under each mode through parallel flow field simulation. Step 3: Establish a surrogate model with the geometric design variables as input and the aerodynamic response data as output; set the torsional moment of the vertical take-off and landing mode and the pitching moment of the attitude transition mode as the primary optimization objective, and set the lift-to-drag ratio of the horizontal cruise mode as the secondary optimization objective, with the primary optimization objective having a higher priority than the secondary optimization objective. Using the geometric design variables as optimization variables, a genetic algorithm is used for iterative optimization, and the values of the optimization variables are limited to not exceeding the spatial range covered by the training samples of the surrogate model, and finally the optimal configuration parameters are output.
2. The aerodynamic optimization design method according to claim 1, characterized in that, In step 1, the value range of the geometric design variables is 0.65m. <D in <0.68m, 0.6m <D out <0.63m, 0.24m <L<0.26m、0.03m<t<0.04m、0.04m< l1<0.06m、0.1m< l2<0.12m、0.14m< c<0.16m。 3. The aerodynamic optimization design method according to claim 1, characterized in that, In step 2, the boundary conditions for the three flight modes are configured as follows: Vertical takeoff and landing mode: incoming flow velocity 0 m / s, total thrust of propeller momentum source 25 kgf, overall angle of attack 0°, all tail rudder surfaces uniformly deflected 5° to counteract the torsional torque generated by propeller counter-torque; Attitude transition mode: incoming flow velocity 10 m / s, total propeller momentum source thrust 25 kgf, overall angle of attack 2°, elevator surfaces additionally deflected downward 5° to provide pitch moment for attitude transition; Horizontal cruise mode: incoming flow velocity 42 m / s, total thrust of propeller momentum source 6 kgf, overall angle of attack 2°.
4. The aerodynamic optimization design method according to any one of claims 1-3, characterized in that, In step 2, the aerodynamic response data includes six components of the whole aircraft data under each mode: lift, drag, side force, roll moment, pitch moment and yaw moment; among them, the torsional moment in the roll direction is extracted in the vertical take-off and landing mode, the pitch moment is extracted in the attitude transition mode, and the lift and drag are extracted in the horizontal cruise mode for calculating the lift-to-drag ratio.
5. The aerodynamic optimization design method according to claim 1, characterized in that, In step 2, the k-ω SST two-equation turbulence model is uniformly selected, and the spatial discretization adopts the second-order upwind scheme; the computational grid adopts a hybrid grid, with hexahedral grids in the boundary layer region and tetrahedral grids in the far-field region. The far-field boundary is 20 times the span from the aircraft surface, and the boundary layer is divided into 30 layers. The height of the first layer of the grid satisfies the wall y + < 1.
6. The aerodynamic optimization design method according to claim 1 or 5, characterized in that, In step 2, the propeller defines a momentum source term in a separately partitioned volume grid region within the computational domain to simulate the momentum injection of the propeller disk into the airflow.
7. The aerodynamic optimization design method according to claim 1, characterized in that, In step 3, the proxy model is established as follows: sample points are extracted in the design space using the Latin hypercube sampling method, and the parallel flow field simulation is performed on each sample point to obtain the corresponding aerodynamic response data, forming a training sample pair. Then, a neural network is used to fit the mapping relationship between the geometric design variables and the aerodynamic response data.
8. The aerodynamic optimization design method according to claim 1, characterized in that, In step 3, the quantitative constraints of the primary optimization objective are: the torsional moment of the vertical take-off and landing mode is ≥8.0 N·m, and the pitching moment of the attitude transition mode is ≥5.0 N·m; the secondary optimization objective is: the lift-to-drag ratio of the entire aircraft in the horizontal cruise mode is improved by ≥4.0% compared with the initial shape.
9. The aerodynamic optimization design method according to claim 1 or 8, characterized in that, In step 3, when comparing the merits of two individuals, it is first determined whether the primary optimization objective has met the quantitative constraints. Individuals that have not met the primary optimization objective are eliminated. Only when both individuals have met the primary optimization objectives are the values of their secondary optimization objectives compared, and the one with the better secondary optimization objective is retained.
10. The aerodynamic optimization design method according to claim 1, characterized in that, In step 3, after outputting the optimal configuration parameters, a verification step is also included: the optimal geometric design variables output by optimization are substituted back into the parallel flow field simulation process in step 2 to independently verify the three-mode aerodynamic response. If the deviation between the verification result and the surrogate model's predicted value exceeds a preset threshold, the sampling step is returned to supplement training samples and the surrogate model is retrained. Iterative optimization is then performed again until the verification passes.