Multidisciplinary Aero‑Structural Optimization of a Wing Based on the PPO Algorithm

The pursuit of fuel-efficient and lightweight aircraft demands advanced aero-structural optimization. However, conventional gradient-based approaches face significant challenges in high-dimensional design spaces characterized by strong nonlinear couplings. Deep reinforcement learning (DRL) offers a promising alternative, yet public research has largely been confined to 2D airfoils or single-discipline problems. This paper presents a DRL framework for the design of 3D wing geometries. By integrating Proximal Policy Optimization (PPO) with the open-source OpenAeroStruct toolchain, the agent learns to adjust the spanwise twist and thickness distributions of the wing to improve aerodynamic efficiency and reduce fuel consumption while satisfying structural constraints. The results show that the proposed framework can identify high-performing and feasible designs under fixed operating conditions. This work establishes a practical baseline for DRL in complex aero-structural optimization.