基于机器学习的船舶水润滑轴承结构多目标优化研究  

Multi-objective Optimization Study of Water-lubricated Bearing Structure for Ships Based on Machine Learning

作　　者：刘辉[1] 于鹏法 陈紫起 LIU Hui;YU Pengfa;CHEN Ziqi(School of Naval Architecture&Ocean Engineering,Dalian University of Technology,Dalian 116024,China)

机构地区：[1]大连理工大学船舶工程学院,大连116024

出　　处：《中国造船》2024年第4期133-144,共12页Shipbuilding of China

基　　金：中国博士后科学基金资助项目(2023TQ0041,2023M7404771);国家资助博士后研究人员计划项目(GZC20230347)。

摘　　要：水润滑轴承是船舶轴系安全稳定运转的重要支撑部件,通过优化轴承结构以提升轴承性能是保证船舶安全航行的有效措施。采用PSO-BP神经网络建立水润滑轴承承载力和摩擦力预测模型,应用NSGA-Ⅱ以预测模型承载力最大和摩擦力最小为优化目标,优化预测模型的输入值使目标函数达到最优,得到轴承的Pareto解,通过TOPSIS方法选取Pareto解集的最优非劣解。结果表明:优化后的轴承承载力较原始设计值提高了38.17%,摩擦力降低了2.23%;预测模型的优化结果与将优化参数输入仿真模型计算得到的结果相比,误差小于7%。Water-lubricated bearing is an important supporting component for the safe and stable operation of ship shaft systems,and it is an effective measure to ensure the safe navigation of ships and improve bearing performance.The PSO-BP neural network is introduced to establish an agent model of load-carrying capacity and friction force of water-lubricated bearings.The NSGA-I algorithm is applied to take the maximum loadcarrying capacity output and the minimum friction force output of the agent model.Then the input of the agent model is optimized,and the Pareto solution of the bearings is obtained.The optimal compromise solution in the Pareto solution set is selected using the TOPSIS method.The results show that the optimized bearing load carrying capacity is increased by 38.17%,and friction force is reduced by 2.23%compared with the original design of the bearing.The percentage difference in optimization results is less than 7%between the agent and the simulation model.

关 键 词：水润滑轴承 结构优化 PSO-BP神经网络 NSGA-Ⅱ TOPSIS方法 

分 类 号：U664.21[交通运输工程—船舶及航道工程]