A physics-informed design framework for deep-sea pressure hulls: decoupling structural-acoustic trade-offs via surrogate modeling

Designing deep-sea pressure hulls involves a fundamental trade-off between structural integrity and acoustic stealth, where decreasing wall thickness to reduce weight often compromises the inertial resistance necessary for optimal sound transmission loss (STL). To overcome the computationally prohibitive cost of iterative finite element evaluations and to facilitate the efficient vibro-acoustic optimization of complex shell geometries, this study proposes a physics-informed surrogate modeling framework. Initial parametric analyses indicate that manipulating geometric stiffness and achieving resonance decoupling provide more effective structural-acoustic isolation mechanisms under high hydrostatic pressure than relying solely on the classical mass law. Subsequently, a deep feedforward neural network (FNN) trained on physics-generated spectral data is integrated with a genetic algorithm (GA) to navigate this high-dimensional design space and decouple these competing physical constraints. The framework identifies an optimized geometric configuration that achieves a mean STL improvement of 7.92 dB across the target frequency band while simultaneously reducing structural weight by 59%. This work demonstrates that mechanical robustness and acoustic stealth can be effectively reconciled, providing a numerically validated computational strategy for the development of low-noise deep-sea submersibles.