Optimising General Configuration of Wing-Sailed Autonomous Sailing Monohulls Using Bayesian Optimisation and Knowledge Transfer
Published in Journal of Marine Science and Engineering, 2023
针对翼帆无人单体帆船设计中 CFD 仿真代价高、迭代效率低的问题，本文提出一种将贝叶斯优化（BO）与知识迁移（KT）相结合的双代理模型框架。传统方案在缺乏先验性能数据时往往需要大量 CFD 计算，而本方法通过 BO 将采样自适应聚焦于关键工况附近，并利用 KT 代理模型复用历史 CFD 数据，从而显著降低计算开销。在“Seagull”原型验证中，在安全约束（横倾角 ≤26°）下实现迎风 56%、顺风 106% 的航速提升；相较离线方法仿真成本降低约 60%。该研究为任务定制化无人帆船的高效设计提供了通用方法，并可推广至水翼耦合等复杂系统。
This research addresses the computational inefficiency in designing wing-sailed autonomous monohulls by introducing a novel dual-surrogate framework combining Bayesian optimisation (BO) and knowledge transfer (KT). Traditional methods require prohibitive CFD simulations due to lack of prior performance data, while the proposed approach dynamically focuses sampling near critical operating points using BO and reuses historical CFD data through KT surrogates. Validated on the “Seagull” prototype, the framework achieved 56% upwind and 106% downwind speed improvements under safety constraints (heel angle ≤26°), demonstrating 60% reduction in simulation costs compared to offline methods. The work provides a generalised, efficient design methodology for mission-specific autonomous sailboats, with potential applications in hydrofoil-coupled systems.
Recommended citation: Yang An*, Feng Hu, Kuo Chen, and Jiancheng Yu*. (2023). "Optimising General Configuration of Wing-Sailed Autonomous Sailing Monohulls Using Bayesian Optimisation and Knowledge Transfer." Journal of Marine Science and Engineering. 11(4), 703.
Download Paper