关键词: 吸气式高超声速飞行器, 推力损失, 在线容错轨迹规划, 神经网络, 燃料消耗预测, 扰动加速粒子群

Abstract: Aiming at the problem of sudden thrust loss during the flight of air-breathing hypersonic vehicles that affects mission objectives and flight safety, this paper proposes an online fault-tolerant trajectory planning method. First, a longitudinal dynamics model and climb profile strategy model for air-breathing hypersonic vehicles are established. The trajectory optimization is transformed into a finite-parameter optimization problem based on the climb profile, and performance metrics for online trajectory planning are analyzed. Sub-sequently, during the offline phase, research on flight capability assessment methods and deep neural network (DNN) modeling for fuel consumption prediction is conducted. By converting state constraints and control constraints into flight capability boundaries, the complexity of constraints in the trajectory optimization problem is resolved. A mapping relationship between flight state change rates and fuel consumption rates is established based on knowledge of the dynamics model, significantly enhancing the efficiency of fuel consumption prediction. During the online phase, trajectory parameters are optimized using a disturbance-accelerated particle swarm optimization algorithm, balancing optimization efficiency and stability. Mathematical simulations demonstrate that the proposed method achieves rapid trajectory parameter optimization within 1 second. Compared with DNN-based optimization methods, this approach reduces fuel consumption by 1.3% and shortens flight time by 3.4%, verifying its effectiveness.

Key words: Air-breathing Hypersonic Vehicle, Thrust Loss, Online Fault-Tolerant Trajectory Planning, Neural network, Fuel Con-sumption Prediction, Disturbance-accelerated Particle Swarm