Sailboat Hull Resistance Dataset and Predictive Model

This document contains a dataset and a predictive model. The dataset is based on three sailboats systematic series with measurement of hull resistance through water: The Delft Series, US Sailing Series and Il Moro di Venezia Series. The data are stored in a Coma Separated Virgula text file (“Sailboat Hull Resistance Dataset V01.csv”) whereas the semicolon has been used as a separator character. The table includes 1018 records corresponding to towing tank tests of the systematic sailboat hull series. Respectively, 702 records are related to the Delft Series, 108 records to the Il Moro di Venezia Series and 208 records to the US Sailing Series. The table possesses 21 fields that are described in detail in the metadata file (“Sailboat Hull Resistance Metadata V01.csv”). The predictive model uses an Artificial Neural Network configured with 8 inputs to predict the hull resistance target variable “Rt/Delta * 10^3”. The inputs of the model are: The Froude number, the Reynolds number and the 8 principal components (PCA) of the following fields: Cp; Cm; Cb; Cw; Lwl/Bwl; Bwl/Tc; Lwl/Tc; Lwl/Vol^(1/3); Lcb/Lwl; Lcf/Lwl; Lcb/Lcf; Sc/Vol^(2/3); Aw/Vol^(2/3); Sc/Aw; Sc/Ax; Ax/Aw. The predictive model uses the PMML 4.2.1 data format (http://dmg.org/pmml/pmml-v4-2-1.html). The model is stored in the pmml file: (“ann_3-16_r2_0.998.pmml”). The neural network has 3 hidden layers of 16 neurons. The details of the step by step procedures to use the predictive model are available in the paper referenced here: https://doi.org/10.1016/j.oceaneng.2022.111642.本文件包含一套数据集与一个预测模型。该数据集基于三类带有船体水下阻力测量数据的帆船系统序列：代尔夫特序列（Delft Series）、美国帆船序列（US Sailing Series）与伊尔·莫罗·迪·威尼斯序列（Il Moro di Venezia Series）。数据以逗号分隔值（Comma Separated Values，CSV）文本文件格式存储，文件名为"Sailboat Hull Resistance Dataset V01.csv"，实际采用分号作为字段分隔符。该表格共包含1018条记录，对应系统帆船船体系列的拖曳水池试验数据，其中代尔夫特序列占702条，伊尔·莫罗·迪·威尼斯序列占108条，美国帆船序列占208条。该表格共设21个字段，详细说明见元数据文件"Sailboat Hull Resistance Metadata V01.csv"。本预测模型采用人工神经网络（Artificial Neural Network）构建，配置8个输入变量，用于预测船体阻力目标变量"Rt/Delta * 10^3"。模型输入包括：傅汝德数（Froude number）、雷诺数（Reynolds number），以及以下字段的8项主成分（PCA）：Cp、Cm、Cb、Cw、Lwl/Bwl、Bwl/Tc、Lwl/Tc、Lwl/Vol^(1/3)、Lcb/Lwl、Lcf/Lwl、Lcb/Lcf、Sc/Vol^(2/3)、Aw/Vol^(2/3)、Sc/Aw、Sc/Ax、Ax/Aw。该预测模型采用PMML 4.2.1数据格式（http://dmg.org/pmml/pmml-v4-2-1.html），存储于pmml文件"ann_3-16_r2_0.998.pmml"中。该神经网络包含3个隐藏层，每层含16个神经元。关于该预测模型的详细分步使用流程，可参见下述参考文献中的论文：https://doi.org/10.1016/j.oceaneng.2022.111642。