The rapid evolution of artificial intelligence (AI) has unlocked transformative pathways for high-performance aerospace engineering, yet its integration into missile and propellant design remains nascent. This study investigates a hybrid AI-driven framework that unites physics-based modeling, data-centric learning, and multi-objective optimization to accelerate the conception, validation, and refinement of missile airframes and solid-propellant formulations. First, a physics-informed neural network (PINN) is trained on a curated database of historic missile geometries, material properties, and flight-test data, enabling the rapid prediction of aerodynamic coefficients and structural loads across a broad design envelope. Second, a generative adversarial network (GAN) produces candidate airframe topologies that satisfy stealth, maneuverability, and thermal-signature constraints while respecting manufacturability rules encoded as conditional priors. Third, a reinforcement-learning (RL) agent iteratively proposes propellant grain morphologies and composite chemistries, receiving reward signals from a high-fidelity thermochemical solver that evaluates specific impulse, burn rate stability, and mechanical integrity. The AI pipeline operates in a closed-loop fashion: surrogate models quickly screen millions of configurations, the RL agent fine-tunes promising candidates, and the most viable designs are passed to a reduced-order CFD/combustion suite for verification. Compared with conventional manual iteration, the proposed methodology reduces total design cycle time by ≈73 %, discovers 12 % higher specific-impulse propellants, and yields airframe shapes that lower radar cross-section by 18 % without sacrificing payload capacity. Sensitivity analyses demonstrate robustness against uncertainties in material aging and launch-environment variability. The findings affirm that AI can transcend its role as a mere optimization tool, becoming an autonomous co-designer that internalizes domain physics, explores unconventional solution spaces, and expedites the translation from concept to flight-ready hardware.

🔒 This is a subscription article

Full text is available to subscribers and institutional members. Please choose an option below to access it.

The rapid evolution of artificial intelligence (AI) has unlocked transformative pathways for high-performance aerospace engineering, yet its integration into missile and propellant design remains nascent. This study investigates a hybrid AI-driven framework that unites physics-based modeling, data-centric learning, and multi-objective optimization to accelerate the conception, validation, and refinement of missile airframes and solid-propellant formulations. First, a physics-informed neural network (PINN) is trained on a curated database of historic missile geometries, material properties, and flight-test data, enabling the rapid prediction of aerodynamic coefficients and structural loads across a broad design envelope. Second, a generative adversarial network (GAN) produces candidate airframe topologies that satisfy stealth, maneuverability, and thermal-signature constraints while respecting manufacturability rules encoded as conditional priors. Third, a reinforcement-learning (RL) agent iteratively proposes propellant grain morphologies and composite chemistries, receiving reward signals from a high-fidelity thermochemical solver that evaluates specific impulse, burn rate stability, and mechanical integrity. The AI pipeline operates in a closed-loop fashion: surrogate models quickly screen millions of configurations, the RL agent fine-tunes promising candidates, and the most viable designs are passed to a reduced-order CFD/combustion suite for verification. Compared with conventional manual iteration, the proposed methodology reduces total design cycle time by ≈73 %, discovers 12 % higher specific-impulse propellants, and yields airframe shapes that lower radar cross-section by 18 % without sacrificing payload capacity. Sensitivity analyses demonstrate robustness against uncertainties in material aging and launch-environment variability. The findings affirm that AI can transcend its role as a mere optimization tool, becoming an autonomous co-designer that internalizes domain physics, explores unconventional solution spaces, and expedites the translation from concept to flight-ready hardware.

🔒 This is a subscription article

Full text is available to subscribers and institutional members. Please choose an option below to access it.