摘要
为了提高船舶上层建筑整体纵向固有频率特性预报的精度,提出一种样本筛选与变权重的船舶上层建筑整体纵向固有频率特性组合预报算法。首先根据样本之间的相似性,选择与待预报点相关的船舶上层建筑整体纵向固有频率特性数据,然后根据采用多个算法从不同角度对船舶上层建筑整体纵向固有频率特性进行预报,最后采用证据理论对单一算法的权值进行估计,加权得到船舶上层建筑整体纵向固有频率特性预报结果,测试结果表明,本文算法的船舶上层建筑整体纵向固有频率特性预报精度高,克服了单一算法或者传组合算法的弊端,降低了船舶上层建筑整体纵向固有频率特性预报误差。
In order to improve the accuracy of forecasting the longitudinal natural frequency characteristics of ship superstructure,a combined forecasting algorithm based on sample selection and variable weight is proposed.Firstly,according to the similarity between samples,the longitudinal natural frequency characteristic data of ship superstructure related to the predicted points are selected,then the longitudinal natural frequency characteristics of ship superstructure are predicted from different angles by using multiple algorithms.Finally,the weight of single algorithm is estimated by using evidence theory,and the longitudinal natural frequency of ship superstructure is obtained by weighting.The test results show that the proposed algorithm has high accuracy in predicting the longitudinal natural frequency characteristics of ship superstructure,overcomes the drawbacks of single algorithm or combined transmission algorithm,and reduces the prediction error of the longitudinal natural frequency characteristics of ship superstructure.
作者
龚洁
GONG Jie(City College of Science and Technology Chongqing University,Chongqing 402167,China)
出处
《舰船科学技术》
北大核心
2019年第12期1-3,共3页
Ship Science and Technology
关键词
船舶上层建筑
纵向固有频率
证据理论
变化特性
组预报算法设计
ship superstructure
longitudinal natural frequency
evidence theory
change characteristics
group prediction algorithm design