強化学習ベースのトポロジー最適化を用いた広帯域音響吸収用水中コーティングの設計

Underwater acoustic coatings play a vital role in minimizing noise and enhancing stealth capabilities by maximizing sound absorption performance. In this study, we introduce a reinforcement learning-based topology optimization (RL-TO) method for designing underwater acoustic coatings that achieve broadband and low-frequency sound absorption. The RL-TO approach integrates reinforcement learning with a systematic topology optimization framework to dynamically explore optimal material distributions and structural configurations. The optimized topologies exhibit distinct characteristics, such as waveguiding at low frequencies and local resonance at high frequencies, demonstrating enhanced absorption performance across a wide frequency range. Further parametric studies on triangular-shaped wave controllers and spherical resonators highlight the importance of geometric features in achieving superior absorption. The proposed method achieves an average absorption coefficient exceeding 0.9 for multiple target frequencies while maintaining robust performance. This study establishes RL-TO as a powerful and efficient design tool for underwater acoustic coatings, offering significant advancements over conventional approaches.