基于PI定理的船舶冲击环境预报方法研究  

A method for predicting ship impact environment based on PI theorem

作　　者：赵晓俊 郭君[1] 杨俊杰 王茀凡 ZHAO Xiaojun;GUO Jun;YANG Junjie;WANG Fufan(College of Shipbuilding Engineering,Harbin Engineering University,Harbin 150001,China;Dalian Shipbuilding Industry Co.,Ltd.,Dalian 116005,China)

机构地区：[1]哈尔滨工程大学船舶工程学院,哈尔滨150001 [2]大连船舶重工集团有限公司,辽宁大连116005

出　　处：《振动与冲击》2023年第19期137-143,共7页Journal of Vibration and Shock

基　　金：国家科技重大专项(J2019-I-0017-0016)。

摘　　要：考虑到船舶水下非接触爆炸冲击响应的强非线性传递规律,采用K-Means聚类以及二分K聚类算法搭建的RBF网络对其冲击响应的谱速度值加以训练和预报。在网络中,选取仿真计算过程中涉及到的诸多参变量来表征船舶冲击环境的特征参数,利用量纲分析及π-定理理论分析方法将原始特征参数转换为无量纲量并且剔除不适应的参量。结果表明,参数无量纲化处理可以提高网络对响应谱速度值的预报准确性,并使预报值偏离仿真结果10%范围内的占比有明显提升,此外得出结合无量纲参数的二分K聚类RBF网络模型对冲击环境谱速度值的预报效果最佳。Here,considering strong nonlinear transmission law of ship underwater non-contact explosion shock response,a RBF network constructed using K-means clustering and bisecting K clustering algorithm was used to train and predict spectral velocity values of the ship’s shock response.In RBF network,many parameters involved in simulation calculation process were selected to characterize characteristic parameters of ship impact environment,and dimensional analysis andπ-theorem theoretical analysis method were used to convert original characteristic parameters into dimensionless quantities and eliminate unsuitable parameters.The results showed that dimensionless parameter processing can improve the correctness of network prediction of response spectral velocity values,and obviously increase the proportion of predicted values deviating from simulation results within 10%range;the bisecting K clustering RBF network model combined with dimensionless parameters has the best prediction effect on impact environmental spectral velocity values.

关 键 词：冲击响应 无量纲 神经网络 聚类 

分 类 号：U661.44[交通运输工程—船舶及航道工程]