Multiphysics modeling and heat transfer enhancement of underwater vehicle thermal engines

Abstract

Thermal engines are essential energy storage components in underwater vehicles that harvest ocean thermal energy. However, most existing studies rely on simplified heat transfer models and fail to capture complex multiphysics interactions. This study develops a comprehensive numerical approach that tracks phase change dynamics, buoyancy-driven convection, and flexible hose deformation simultaneously. To overcome convergence difficulties in large-scale simulations, we developed a physics-constrained gated recurrent unit (GRU) neural network with temporal correction to predict liquid fraction evolution, which can scale results from small-scale simulations (100–250 mm) to full-size prototypes (1100 mm). Experimental validation demonstrates excellent agreement with predictions, achieving a root mean square error of 0.0516 for liquid fraction. Using this validated framework, we investigated how radial fins enhance heat transfer. Results indicate that radial fins reduce the melting time of phase change material (PCM) by 29.4 %, with a 17.6 % improvement in heat transfer area and a 14.3 % enhancement in convection. Among different fin orientations, horizontal fins (0°) are the most efficient of all the fin orientations. They cut melting time by 22 % at a 95 % liquid fraction compared to the -45° orientation. For T-shaped fins, extending the vertical bar from 1 mm to 16 mm results in just an 8–13 % decrease in melting time, even though the volume increases by 16 times, indicating considerable diminishing returns. This paper offers theoretical insights and practical directions for the design of thermal engines in ocean energy applications.