Xian-Jun He

The propeller design for multi-rotor hybrid aquatic–aerial vehicles is faced with a critical trade-off between aerial and aquatic efficiency. To address this challenge, a surrogate model based on multi-fidelity neural networks and transfer learning is proposed to predict aerodynamic and hydrodynamic performance (thrust and torque coefficients) of propellers. The chord length and the pitch angle distribution functions, each governed by three planform variables, are defined to determine blade geometry. Low-fidelity samples and high-fidelity samples derive from blade element momentum theory and computational fluid dynamics simulations, respectively. The non-dominated sorting genetic algorithm II seeks Pareto-optimal solutions through multi-objective optimization. During optimization, thrust and torque constraints are imposed. The optimization seeks to maximize aquatic efficiency without compromising aerial efficiency. The Quantitative analysis compares the aerodynamic and hydrodynamic performance of baseline vs optimized propellers. The optimized propeller achieves substantially higher aquatic thrust and efficiency while maintaining modest aerial thrust and efficiency improvements compared to the baseline propeller, t. Specifically, aerial thrust increases from 0.1198 to 0.1205, with efficiency rising from 17.08% to 17.18%. Aquatic thrust coefficient increases from 0.0372 to 0.0424, with efficiency rising from 63.25% to 69.43%.

To address the complex coupling between aerodynamic characteristics and guidance control for morphing flight missiles, this study proposes a data‐driven approach to integrated adaptive morphing and guidance. Firstly, an aerodynamic surrogate model is constructed using a fully connected neural network (FCNN), mapping the configuration parameters to aerodynamic parameters. Secondly, an adaptive physical parameters optimization network (PPON) is developed to optimize aerodynamic characteristics based on predictions from the aerodynamic surrogate model. Thirdly, an integrated morphing and guidance model is derived by applying the proximal policy optimization (PPO) algorithm from deep reinforcement learning (DRL), embedded with the adaptive aerodynamic optimization model. Eventually, the proposed integrated approach is applied to the guidance task of a morphing cruise missile with variable camber wings. Simulation results demonstrate that the integrated guidance model significantly enhances aerodynamic performance and generates more continuous guidance commands within approximately 4.3 s, outperforming the deep Q‐Network (DQN) algorithm under morphing flight conditions. Moreover, compared to the PPO and DQN‐based guidance laws without morphing flight conditions, the integrated model improves both the guidance accuracy and terminal kinetic energy. Furthermore, the integrated guidance model, trained on stationary targets, remains effective for engaging moving and maneuvering targets, showcasing its robust generalization capability.

The flow dynamics around tandem cylinders are complex, with significant engineering implications, especially in applications like high-rise buildings. This study presents a jet flow control framework for two tandem cylinders with a Reynolds number (Re) of 100, based on deep reinforcement learning. We compare two control strategies: (1) a single-agent strategy, where one controller manages the jet flow for two cylinders and (2) a dual-agent strategy, where separate controllers regulate each cylinder independently. The effectiveness of both strategies is evaluated under varying cylinder radii and inter-cylinder spacing. The results show that the single-agent strategy achieves drag reductions of approximately 28% and 40% for the front and rear cylinders, respectively, while the dual-agent strategy results in reductions of about 32% and 31%. While the single-agent strategy is more effective at reducing drag on the rear cylinder, the dual-agent strategy provides superior drag reduction for the larger cylinder and exhibits smaller fluctuations in drag across all conditions.

Computational efficiency and precision pose a classic contradiction in aerodynamic shape optimization. To address this challenge, this study introduces an effective optimization framework based on multi-fidelity fully connected neural network (MFFCN). The framework utilizes transfer learning (TL) to train a multi-fidelity surrogate model that establishes direct mappings between geometric configuration parameters and aerodynamic performance by adaptively capturing linear or nonlinear relationships concealed between high-fidelity (HF) and low-fidelity (LF) information. The HF and LF data are derived from fine and coarse grids, respectively, evaluated using the same computational fluid dynamics (CFD) model. The MFFCN-TL framework is applied to optimize the National Advisory Committee for Aeronautics 0012 (NACA0012) airfoil (12 design variables) and the Office National d'Études et de Recherches Aérospatiales M6 (ONERA M6) wing (50 design variables). Simulation results demonstrate that the NACA0012 airfoil achieves a 69.47% enhancement in lift–drag ratio in 1.069 s, compared to a 12.76% gain over 24.8 h in single-fidelity CFD-based optimization. The ONERA M6 wing achieves a 24.66% reduction in drag coefficient in 694 ms compared to 18.37% over 237.3 h in the CFD model. Statistical results show that the MFFCN-TL framework can reduce optimization cost by more than 90% compared to the single-fidelity CFD-based model. These findings suggest that the MFFCN-TL framework significantly enhances optimization efficiency and provides superior feasible solutions over single-fidelity methods.