文章摘要
基于神经网络的船舶稳性预报研究
Research on Ship Stability Performance Prediction Based on Neural Network
投稿时间:2022-06-12  修订日期:2022-07-26
DOI:
中文关键词: RBF神经网络  船舶稳性预报  第二代完整稳性  失效概率  输入特征选取
英文关键词: RBF neural network  prediction of the ship stability performance  second generation intact stability  failure probability  determination of the input features
基金项目:国家自然科学基金项目(面上项目,重点项目,重大项目)
作者单位
蒋柴丞 大连理工大学船舶工程学院 
李楷 大连理工大学船舶工程学院 
马坤 大连理工大学船舶工程学院 
摘要点击次数: 66
全文下载次数: 0
中文摘要:
      为省略船舶稳性性能计算中横摇阻尼以及船舶阻力等参数的复杂计算过程,提出一种采用RBF神经网络对第二代完整稳性失效概率预报的方法。研究包括:过度加速度、瘫船稳性以及骑浪/横甩三种失效模式。通过研究船舶相关参数对各失效模式失效概率的影响,确定采用神经网络对每种失效模式进行预报时输入特征的选取,从而降低训练时长。使用训练后的网络对样本船稳性性能进行预报,采用均方误差和平均绝对百分误差对预报结果进行评估。对三种失效模式的预报结果,平均绝对百分比误差分别为6.49、7.09、8.49,均小于10,表明采用RBF神经网络可较为精准的对船舶稳性性能进行预报。
英文摘要:
      In order to avoid the complicated calculation process of roll damping and ship resistance in the calculation of ship stability performance, a method for predicting the failure probability of the second generation intact stability by using Radial Basis Function (RBF) neural network was proposed. In this study, three failure modes were included: excessive acceleration, dead-ship stability, and surfing riding/broaching. By studying the influence of ship-related parameters on the failure probability of each failure mode, the input features when using neural network to predict each failure mode were determined, thereby reducing the training time. The trained network was used to predict the stability performance of the sample ship, and the Mean Square Error and Mean Absolute Percentage Error (MAPE) were used to evaluate the predicted values. For the prediction values of the three failure modes, the MAPE are 6.49, 7.09, and 8.49, all of which are less than 10, indicating that the RBF neural network can be used to predict the ship stability performance more accurately.
View Fulltext   查看/发表评论  下载PDF阅读器
关闭