基于机器学习的船舶螺旋桨敞水性能预报代理模型  

Prediction of Open-Water Characteristics of Ship Propellers Based on Machine Learning Surrogate Model

作　　者：强以铭 陈诗楠 陈奕宏[1] 褚学森[1] QIANG Yiming;CHEN Shinan;CHEN Yihong;CHU Xuesen(China Ship Scientific Research Center,Wuxi 214082,China;Xi'an Jiaotong University,Xi'an 710049,China)

机构地区：[1]中国船舶科学研究中心,无锡214082 [2]西安交通大学,西安710049

出　　处：《中国造船》2022年第5期181-188,共8页Shipbuilding of China

摘　　要：介绍了基于机器学习的经典模型,以螺旋桨特征参数为输入,建立船舶螺旋桨敞水性能预报代理模型。对几种典型的回归模型,分析了它们的预报精度和适用性。用随机森林方法建立了船舶螺旋桨敞水性能预报代理模型,其预报结果与试验数据吻合较好。对未知桨试验结果的验证表明,KT和10KQ的预报偏差在5%以内,敞水效率预报偏差在2%以内。该研究为船舶螺旋桨性能预报提供了新的手段,有望缩短设计周期,提高设计效率。This paper builds surrogate models to predict open-water characteristics of ship propellers based on classic machine learning theories with propeller geometry features as the input. The paper analyzed the prediction accuracy and adaptability for classic regression models. Random forest, as one of the proposed models, has good prediction accuracy. For a testing propeller, the prediction of KT and 10KQ have less than 5% error, and the open-water efficiency has less than 2% error. This study provides a novel approach to predict characteristics of ship propellers, and reduce the preliminary design life cycle, as well as improve the design efficiency.

关 键 词：机器学习 船舶螺旋桨 敞水性能 随机森林模型 

分 类 号：U661.313[交通运输工程—船舶及航道工程] U662[交通运输工程—船舶与海洋工程]