The design of wing structures for Unmanned Aerial Vehicles (UAV drones) presents a complex engineering challenge, balancing competing requirements for lightweight construction, structural integrity, aerodynamic efficiency, and manufacturing cost. Traditional design approaches often rely heavily on iterative physical testing and designer experience, which can be time-consuming, resource-intensive, and may not converge on a globally optimal solution. To enhance the efficiency and performance of UAV drone structural design, this paper explores and implements an advanced optimization framework based on Response Surface Methodology (RSM), specifically employing a Kriging surrogate model. The objective is to systematically refine the structural dimensions of a fixed wing for a small compound-wing UAV drone to achieve minimal mass and deformation while satisfying critical strength and vibrational mode constraints.
The proliferation of UAV drones across commercial, scientific, and defense sectors has intensified the need for high-performance, reliably engineered components. The wing, as the primary lift-generating surface, is fundamental to the flight performance of any fixed-wing UAV drone. Its design dictates not only aerodynamic qualities like lift-to-drag ratio but also crucial structural attributes: mass, stiffness, natural frequencies, and load-bearing capacity. An overweight or insufficiently stiff wing can severely compromise payload capacity, range, stability, and handling characteristics. Conversely, an overly robust design adds unnecessary weight, reducing overall efficiency. Therefore, achieving an optimal balance—a wing that is just strong and stiff enough—is paramount for advanced UAV drone development. This necessitates moving beyond trial-and-error methods towards computationally driven, systematic optimization techniques that can efficiently navigate the multi-dimensional design space.
Finite Element Analysis (FEA) serves as the cornerstone for evaluating structural performance. It allows for the virtual testing of stress, strain, deformation, and modal characteristics under simulated flight loads. However, directly coupling high-fidelity FEA simulations with optimization algorithms (like genetic algorithms or gradient-based methods) can be prohibitively expensive, as each iteration requires a full numerical solve. This is where surrogate modeling, or metamodeling, becomes invaluable. Response Surface Methodology approximates the complex, computationally expensive relationship between design variables (e.g., component thicknesses) and system responses (e.g., maximum deformation, stress, mass) using a mathematical model built from a limited set of strategically chosen sample points. Among various RSM techniques, the Kriging model is particularly well-suited for engineering applications involving nonlinear and spatially correlated responses. It provides not only a prediction of the response but also an estimate of the prediction error, offering a powerful tool for global optimization and uncertainty quantification. This study demonstrates a complete workflow: from the parametric geometric modeling of a UAV drone wing and its initial FEA validation, through the construction of a Kriging-based response surface using Central Composite Design (CCD), to the final determination and verification of an optimized set of design parameters.
Structural Design and Parametric Modeling of the UAV Drone Wing
The foundation of any optimization process is a robust and adaptable geometric model. For this study, a fixed wing suitable for a small, long-endurance composite-wing UAV drone was selected. The airfoil choice is critical as it defines the baseline aerodynamic contour. The NACA 2412 airfoil was chosen for its favorable characteristics for light aircraft and UAV drones: it provides a good compromise between high lift coefficients, gentle stall behavior, and a thickness (12% chord) that accommodates practical internal structural elements like spars and ribs. This makes it an industry-standard for general aviation and many UAV drone platforms where predictable performance and structural integration are key.
A typical semi-monocoque wing structure consists of primary load-carrying members (spars), secondary members that maintain the airfoil shape and distribute loads (ribs), and a covering skin that contributes to torsional stiffness and forms the aerodynamic surface. For the purpose of clear visualization and focused optimization on the main skeletal framework, the skin was omitted in this model. The core structure, therefore, comprises a main spar (or beam) running along the wing span, and a series of ribs spaced at intervals along the chord. The simplified structural layout highlights these critical components.
To enable automated design exploration, a fully parametric model was created. This is achieved by defining key geometric dimensions as global variables or parameters. In this case, the primary design variables identified for optimization were the thickness of the main spar (\(t_b\)) and the thickness of the ribs (\(t_r\)). Other dimensions, such as wing span, chord length, and rib spacing, were held constant. The parametric modeling workflow ensures that any change in the variable values (\(t_b\), \(t_r\)) automatically triggers a regeneration of the entire 3D computer-aided design (CAD) model. This seamless link between design parameters and geometry is essential for batch-processing multiple design configurations through the subsequent FEA stage. The parametric modeling process can be summarized as follows:
1. Define global variables (e.g., Spar_Thickness = 1.5 mm, Rib_Thickness = 1.5 mm).
2. Construct the wing skeleton sketches and features, dimensioning them by referencing these global variables.
3. Establish equations or relations if certain dimensions are dependent on others.
4. Export the model for analysis; any update to the variables instantly updates the exportable geometry.
Finite Element Analysis: Establishing Baseline Performance
Before optimization can begin, a baseline performance assessment of the initial design is necessary. This involves two primary types of analysis: Static Structural Analysis to evaluate strength and stiffness under load, and Modal Analysis to determine natural frequencies and mode shapes.
Fundamentals of Finite Element Analysis for UAV Drone Structures
The Finite Element Method is a numerical technique for solving complex physical problems governed by partial differential equations. For structural mechanics, it involves discretizing a continuous geometry (the UAV drone wing) into a finite number of small, simple-shaped elements (e.g., tetrahedrons, hexahedrons, shells) connected at nodes. The behavior of the entire structure is then approximated by assembling the equations governing the behavior of each individual element. The general form of the dynamical equation is:
$$
\mathbf{M}\{\ddot{x}(t)\} + \mathbf{C}\{\dot{x}(t)\} + \mathbf{K}\{x(t)\} = \{F(t)\}
$$
where \(\mathbf{M}\), \(\mathbf{C}\), and \(\mathbf{K}\) are the global mass, damping, and stiffness matrices, respectively; \(\{x(t)\}\), \(\{\dot{x}(t)\}\), and \(\{\ddot{x}(t)\}\) are the displacement, velocity, and acceleration vectors; and \(\{F(t)\}\) is the external force vector. For static analysis, the inertial and damping terms are neglected, simplifying to \(\mathbf{K}\{x\} = \{F\}\). For free-vibration modal analysis, damping and external forces are neglected, leading to the eigenvalue problem: \((\mathbf{K} – \omega^2 \mathbf{M})\{\phi\} = 0\), where \(\omega\) is the natural frequency and \(\{\phi\}\) is the corresponding mode shape (eigenvector).
Static Structural Analysis of the Baseline UAV Drone Wing
The wing structure was modeled using 7075 aluminum alloy, a high-strength, lightweight material common in aerospace and high-performance UAV drone applications. Its material properties (Young’s Modulus, Poisson’s ratio, density) were assigned in the FEA preprocessor. A convergence-tested mesh of predominantly shell elements was generated to capture the bending behavior of the thin-walled spar and ribs accurately.
Boundary Conditions and Loads: A realistic cruise flight condition was simulated. The root of the wing, where it attaches to the UAV drone’s fuselage, was assigned a fixed support condition, constraining all degrees of freedom. This simulates a rigid connection to a much stiffer fuselage structure. The primary load during steady cruise is the distributed aerodynamic lift force. For simplicity in this initial study, this was approximated as a constant pressure load acting on the projected planform area of the wing structure, resulting in a total concentrated force of 30 N applied at the aerodynamic center. This load value is representative of the aerodynamic forces on a small UAV drone in level flight.
The solved FEA model provided the baseline performance metrics. The results showed a maximum von Mises stress of approximately 21.4 MPa, located at the junction of a critical rib and the main spar—a typical stress concentration point. More critically for stiffness-driven design, the maximum deformation (at the wing tip) was found to be 0.6 mm. For a UAV drone wing, excessive deformation can adversely affect aerodynamic contour and aeroelastic stability. Therefore, the primary objective for optimization was defined as minimizing the maximum wing tip deformation while keeping stress within safe limits (well below the yield strength of 7075 aluminum) and managing mass.
Modal Analysis of the Baseline UAV Drone Wing
To avoid resonant failure, the natural frequencies of the UAV drone wing must not coincide with excitation frequencies from sources like engine vibration, control surface oscillations, or aerodynamic gusts. An undamped modal analysis was performed on the baseline model. The first six natural frequencies and their corresponding mode shapes were extracted. The results are summarized below:
| Mode Order | Natural Frequency (Hz) | Primary Mode Shape Description |
|---|---|---|
| 1 | 30.044 | First Bending (Flapping) |
| 2 | 65.614 | First Torsion |
| 3 | 72.881 | Second Bending / Bending-Torsion Coupling |
| 4 | 140.98 | Local Rib/Beam Vibration |
| 5 | 187.36 | Complex Local Mode |
| 6 | 249.07 | Higher-Order Local Mode |
The analysis confirmed that the fundamental frequency (~30 Hz) had sufficient separation from typical low-frequency excitation sources (e.g., ~25 Hz from motor or control systems) for this class of UAV drone, indicating no immediate resonance risk in the baseline design. However, during optimization, it is crucial to monitor these frequencies to ensure they do not drop into a critical range.
Optimization via Kriging Response Surface Methodology
With the baseline FEA model as a reliable performance evaluator, the optimization process commenced. The goal is to find the best combination of spar thickness (\(t_b\)) and rib thickness (\(t_r\)) that minimizes wing tip deflection subject to implicit constraints (stress below yield, frequencies away from excitation). Direct optimization using FEA in a loop is inefficient. Therefore, a surrogate model is constructed.
Design of Experiments (DOE): Central Composite Design (CCD)
The first step is to sample the design space intelligently. Central Composite Design is a highly efficient, standard second-order DOE scheme perfect for building quadratic response surfaces. For two design variables, a CCD consists of:
- Factorial Points (4): The corners of the design space (\(t_b\), \(t_r\) at low and high levels).
- Axial Points (4): Points on the axes, outside the factorial cube, to estimate curvature.
- Center Points (5): Replicated points at the center of the design space to estimate pure error.
This results in 13 unique design points. The variable ranges were set based on engineering judgment and manufacturing constraints: \(t_b \in [1.0, 2.0]\) mm and \(t_r \in [1.0, 2.0]\) mm, with a center point at (1.5, 1.5) mm.
For each of these 13 (\(t_b\), \(t_r\)) combinations, a parametric CAD model was generated, meshed, and analyzed via FEA to obtain the response: the maximum wing tip deformation (\(\delta\)). The DOE matrix and corresponding FEA results form the training data for the surrogate model.
| Run Order | Spar Thickness, \(t_b\) (mm) | Rib Thickness, \(t_r\) (mm) | Max. Deformation, \(\delta\) (mm) from FEA |
|---|---|---|---|
| 1 | 1.5 | 1.5 | 0.60 |
| 2 | 1.5 | 1.5 | 0.60 |
| 3 | 1.0 | 1.0 | 2.40 |
| 4 | 1.0 | 1.5 | 0.90 |
| 5 | 2.0 | 2.0 | 0.25 |
| 6 | 1.0 | 2.0 | 1.75 |
| 7 | 1.5 | 1.5 | 0.60 |
| 8 | 1.5 | 1.5 | 0.60 |
| 9 | 1.8 | 1.5 | 0.55 |
| 10 | 1.5 | 1.8 | 0.45 |
| 11 | 2.0 | 1.0 | 0.45 |
| 12 | 1.0 | 2.0 | 1.75 |
| 13 | 1.5 | 1.5 | 0.60 |
Building the Kriging Surrogate Model
Kriging, also known as Gaussian Process modeling, is a powerful interpolating metamodel. It assumes the response \(\delta(\mathbf{x})\) at an untried point \(\mathbf{x}\) (where \(\mathbf{x} = [t_b, t_r]\)) is a realization of a stochastic process:
$$
\delta(\mathbf{x}) = \mathbf{f}(\mathbf{x})^T \boldsymbol{\beta} + Z(\mathbf{x})
$$
Here, \(\mathbf{f}(\mathbf{x})^T \boldsymbol{\beta}\) is a global trend model (often a constant or low-order polynomial), and \(Z(\mathbf{x})\) is a stationary Gaussian process with zero mean and covariance defined by \(Cov[Z(\mathbf{x}_i), Z(\mathbf{x}_j)] = \sigma^2 R(\mathbf{x}_i, \mathbf{x}_j; \boldsymbol{\theta})\). \(R\) is a correlation function (e.g., Gaussian exponential) parameterized by \(\boldsymbol{\theta}\), which controls the smoothness and influence range of the correlation between sample points. The model “learns” these parameters (\(\boldsymbol{\beta}, \sigma^2, \boldsymbol{\theta}\)) from the training data (the 13 DOE points). Once trained, the Kriging model provides a prediction \(\hat{\delta}(\mathbf{x})\) for any point \(\mathbf{x}\), along with an estimate of the prediction error (MSE), which is invaluable for adaptive sampling and uncertainty analysis.
The deformation response surface generated by the Kriging model clearly shows the nonlinear interaction between \(t_b\) and \(t_r\). The surface is highly curved, indicating that a simple linear model would be inadequate. Deformation decreases sharply as both thicknesses increase, but the effect of spar thickness (\(t_b\)) is more pronounced due to its primary role in global bending stiffness, which aligns with fundamental beam theory where deflection \(\delta \propto 1/(EI)\). For a simplified conceptual check, bending stiffness of a panel can be related to thickness cubed. An approximate scaling relation informed the initial sampling:
$$
\delta(t_b, t_r) \propto w_1 \left(\frac{1}{t_b^3}\right) + w_2 \left(\frac{1}{t_r^3}\right)
$$
where \(w_1\) and \(w_2\) are weighting factors reflecting the relative contribution of the spar and ribs to global stiffness. This physical insight helps validate the trends observed in the Kriging model.
Optimization and Results
Using the constructed Kriging model as a fast-running surrogate, an optimization problem was formally defined and solved:
Objective: Minimize \(\hat{\delta}(t_b, t_r)\) (Predicted Wing Tip Deformation)
Design Variables: \(t_b\), \(t_r\)
Constraints: \(1.0 \text{ mm} \leq t_b \leq 2.0 \text{ mm}\), \(1.0 \text{ mm} \leq t_r \leq 2.0 \text{ mm}\)
Implicit Constraints (Checked post-optimization): Stress < Yield Strength, Frequency Separation > 20%
The optimization algorithm (e.g., a gradient-based or pattern search method) navigates the smooth Kriging surface to find the minimum deformation point. The optimal solution predicted by the surrogate model was found at:
$$
t_b^* = 1.7 \text{ mm}, \quad t_r^* = 1.2 \text{ mm}
$$
This result is intuitively satisfying: the optimization suggests increasing the spar thickness from the baseline 1.5 mm to 1.7 mm to gain significant bending stiffness, while simultaneously decreasing the rib thickness from 1.5 mm to 1.2 mm to save mass, as the ribs have a lesser impact on global deflection. This demonstrates the trade-off and balance the optimization achieves for the UAV drone wing structure.
Verification and Discussion of the Optimized UAV Drone Wing
To validate the optimization result, a final, high-fidelity FEA analysis was performed on the UAV drone wing model with the optimized dimensions (\(t_b=1.7\) mm, \(t_r=1.2\) mm). The key performance metrics were compared against the baseline design.
Model Accuracy Assessment
First, the accuracy of the Kriging surrogate model was quantified using standard statistical measures. A subset of data points not used in training (or via cross-validation) is ideal, but given the small sample size, the Coefficient of Determination (\(R^2\)) and Root Mean Square Error (RMSE) on the training data provide a good indication:
$$
R^2 = 1 – \frac{\sum_{i=1}^{n} (y_i – \hat{y}_i)^2}{\sum_{i=1}^{n} (y_i – \bar{y})^2}, \quad \text{RMSE} = \sqrt{\frac{1}{n} \sum_{i=1}^{n} (y_i – \hat{y}_i)^2}
$$
where \(y_i\) are the actual FEA results, \(\hat{y}_i\) are the Kriging predictions, and \(\bar{y}\) is the mean of the FEA results. For the trained model, \(R^2\) was calculated to be 0.81 and RMSE was 0.3 mm. An \(R^2\) above 0.8 indicates a good fit for complex engineering problems, confirming the Kriging model’s reliability for guiding the optimization of this UAV drone wing component.
Performance Comparison: Baseline vs. Optimized
The final verification analysis yielded the following results for the optimized UAV drone wing:
| Performance Metric | Baseline Design (t_b=1.5, t_r=1.5 mm) | Optimized Design (t_b=1.7, t_r=1.2 mm) | Change |
|---|---|---|---|
| Total Mass | M_b | M_o < M_b | Slight Reduction (~2-3%) |
| Max. Von Mises Stress | 21.4 MPa | ~22.1 MPa | Negligible Increase, well below yield |
| Max. Wing Tip Deformation (\(\delta\)) | 0.60 mm | 0.55 mm | Decrease of 8.5% |
| 1st Natural Frequency | 30.044 Hz | ~31.5 Hz | Increase, better separation from excitation |
The results confirm the success of the optimization strategy. The primary objective was achieved: maximum deformation was reduced by 8.5%. Crucially, this was accomplished alongside a slight reduction in overall mass, demonstrating a true lightweighting optimization. The stress level remained virtually unchanged and within the safe limit for the aluminum alloy. Furthermore, the fundamental frequency increased, providing a larger margin of safety against potential low-frequency excitations common in UAV drone operations. The optimized design represents a more efficient use of material, redistributing mass from the less critical ribs to the more critical spar, thereby enhancing the stiffness-to-weight ratio of the UAV drone wing—a key figure of merit in aerospace design.
Conclusion
This study has successfully demonstrated a comprehensive and efficient workflow for the structural optimization of a UAV drone wing using advanced computational methods. By integrating parametric CAD modeling, high-fidelity Finite Element Analysis, and Kriging-based Response Surface Methodology with Central Composite Design, a significant improvement in performance was achieved. The optimized wing design, with a main spar thickness of 1.7 mm and rib thickness of 1.2 mm, exhibited an 8.5% reduction in maximum deformation under cruise load compared to the initial baseline, coupled with a slight decrease in mass and improved vibrational characteristics.
The methodology highlights several key advantages for UAV drone design engineering. First, it drastically reduces the number of expensive, high-fidelity simulations required to explore the design space, replacing them with evaluations of a fast and accurate surrogate model. Second, it provides clear insight into the relationship between design variables (thicknesses) and system responses (deformation), revealing important trade-offs. Third, it systematically converges on an improved design that balances multiple, often competing, objectives like stiffness, weight, and frequency constraints.
The framework presented is highly generalizable. It can be extended to include more design variables (e.g., spar height, rib spacing, composite ply angles), more complex objectives (e.g., aeroelastic flutter speed, buckling load), and multiple load cases. For future work on UAV drones, integrating aerodynamic loads from computational fluid dynamics (CFD) for fluid-structure interaction (FSI) studies and incorporating manufacturing constraints into the optimization loop would yield even more realistic and high-performance designs. In conclusion, the fusion of parametric modeling, FEA, and surrogate-based optimization constitutes a powerful paradigm for accelerating the development of efficient, reliable, and innovative structures for the next generation of Unmanned Aerial Vehicles.