基于机器学习的翼型水动力性能优化设计  

Optimization Design of Airfoil Hydrodynamic Performance Based on Machine Learning

作　　者：李仰建 李子如 刘谦[1,2] 贺伟[1,2] LI Yangjian;LI Ziru;LIU Qian;HE Wei(Key Laboratory of High Performance Ship Technology(Wuhan University of Technology),Ministry of Education,Wuhan 430063,China;School of Naval Architecture,Ocean and Energy Power Engineering,Wuhan University of Technology,Wuhan 430063,China)

机构地区：[1]高性能船舶技术教育部重点实验室(武汉理工大学),武汉430063 [2]武汉理工大学船海与能源动力工程学院,武汉430063

出　　处：《中国造船》2024年第1期176-189,共14页Shipbuilding of China

基　　金：国家自然科学基金重点国际合作研究项目(51720105011)。

摘　　要：基于机器学习的翼型几何优化设计方法可有效避免复杂的计算流体力学数值求解过程,具有更高的计算效率。对翼型进行参数化表示,构建机器学习模型与优化算法进行学习和预测,能极大地减少翼型优化设计时间。论文开展了基于机器学习的翼型水动力性能预测和优化设计研究。运用CST方法对翼型进行参数化表示;采用XGBoost建立翼型水动力特性快速预报模型;结合机器学习方法和遗传算法,综合考虑升力系数、阻力系数和翼型表面压力系数建立优化模型,完成了某翼型的优化设计与水动力性能分析。结果表明:提出的翼型优化设计方法可获取优良翼型,对船用螺旋桨叶剖面设计优化具有重要意义。Machine learning-based airfoil geometry optimization design methods can effectively avoid complex numerical solution processes and have high efficiency.By parameterizing the airfoil geometry and establishing machine learning models together with optimization algorithms for both prediction and learning,the time required for optimizing airfoil design can be greatly reduced.This paper researches machine learning-based airfoil hydrodynamic performance prediction and optimization design.The CST method is used to parameterize the airfoil,and the ensemble learning method XGBoost is used to establish a rapid forecasting model of airfoil hydrodynamic characteristics.Combined with machine learning methods and genetic algorithms,an optimization model is established.The results show that the proposed airfoil optimization and design method can efficiently obtain optimal airfoil geometries.It has important implications for the sectional geometry design optimization of marine propeller blades.

关 键 词：遗传算法NSGA-Ⅲ 翼型设计 CST参数化 应用软件Open FOAM 

分 类 号：U664.33[交通运输工程—船舶及航道工程]