# Simulated and predicted pressure fields of a container vessel in regular sea-states

v1.0 (02 May 2025)
DOI: https://doi.org/10.15480/882.15084

Authors: Malte Loft (0000-0002-2820-3761), Henning Schwarz (0000-0002-2820-3761), Thomas Rung (0000-0002-3454-1804)
Institute for Fluid Dynamics and Ship Theory, Hamburg University of Technology, Am Schwarzenberg-Campus 4, D-21073 Hamburg, Germany
Corresponding E-Mail: malte.loft@tuhh.de

Software used: 
CFD simulations: In-House code FreSCo+
Postprocessing: Python3.12

## Motivation 

In merchant shippping, routing optimization systems represent an essential pillar to reduce the ecological foot print, e.g. to reduce added resistances due to seaways and/or wind. Accurate predictions of these pressure fields are of growing interest for both industry and academia. 
This data set is generated to provide a suffiecient database of time-resolved pressure fields around a free-floating container vessel for training of a CAE/NN based surrogate model. Both CFD data as well as model predictions are available.  

## Parameter spectrum 

This data set contains 150 interpolated CFD simulation results of a (KRISO) container vessel in various sea-states and 18 corresponding predictions of a CAE/NN surrogate model. 3 different operational parameters are used: wave length lambda, wave height 2a and incidence angle beta of the waves. The parameter spectrum reads 

beta = [0, 20, 40, 60, 80, 100, 120, 140, 160, 180] deg,
lambda = [0.50 0.75 1.00 1.25 1.50] Lpp, 
2a = [1/200 1/100 1/66] Lpp.

An angle of beta = 0 indicated waves encountering the vessel from behind. 3 different CFD grids were used to generate the data. Therefore, 3 addtional calm water simulations were performed to extract pressure difference for the training procedure of the CAE/NN model.  

The full CFD data sets are available from the corresponding author upon reasonable request.

## Notation used 

A brief notation is defined in which a letter denotes the angle beta [a-j] followed by two digits, i.e. one for the wave length [1-5] and one for the height of the waves [1-3], e.g.:

a        1      -       1
L ..... corresponds to the indicence angle beta [a, ..., j] =! [0, ..., 180] deg 
         L ..... corresponds to the wave length [1, ..., 5] =! [0.5, ..., 1.50] Lpp
                        L ..... corresponds to the wave height [1, 2, 3] =! [1/200, 1/100, 1/66] Lpp

The calm water simulations are indicated with cw-1 to cw-3. 

## Interpolation properties 

For each simulation case, the CFD data are interpolated on two different meta grids A and B via an inverse distance approach. The grid properties are shown below: 

Grid A: 128 x 32 x 48 (x_1, x_2, x_3),
Grid B: 128 x 32 x 48 (x_1, x_2, x_3).

Both grids only differ in the x_2 and x_3 resolution and cover the same domain

x_1 = [-0.727871, 8.006581], 
x_2 = [-0.713307, 0.713307],
x_3 = [-0.120257, 0.781671]. 

Values within the ship geometry are set to NaN.


## Predictions by the CAE/NN model 

To also account for angles of incidence in the range 200 deg ≤ beta ≤ 360 deg , the pressure fields for the nine angles 0 deg ≤ beta ≤ 160 deg are mirrored, i.e., pressure fields for beta = 20 deg are flipped along the x_2 axis to obtain pressure fields for beta = 340 deg. The dataset used by the ML procedure therefore contains a total of 285 cases. In total, the following 18 cases were predicted by a CAE/NN surrogate model (m indicates, data were flipped): 

a2-2, b2-1, b2-3m, c3-1m, c4-1, d1-2m, d3-2, c1-3m, c3-3, f5-1, f5-3m, g2-2, g3-2m, h1-3m, h3-1, i3-1m, i5-2, j4-3. 

For these simulations, the data set contains an additional precitions.dat file for both meta grids A and B. 


## Read data in python 

The .dat files can be read into a 3D matrix, e.g. with numpy: 

field = numpy.loadtxt(a1-1/grid_A/mean.dat).reshape(32,128,48)


## Overall data structure 

Each simulation contains 501 instantaneous and interpolated data snapshots of both meta grids A and B as well as a mean pressure field each. Time steps considered for the mean pressure fields are indicated in an additional README.md file. This file also contains information about the corresponding calm water simulation (cw-1, cw-2 or cw-3).

The data structure is as follows: 

Dataset
L ..... README.md
L ..... cw-1
    L ..... README.md
    L ..... grid_A
	    L ..... mean.dat
		L ..... time_series
			L ..... 10000.dat
			L ..... 10020.dat
			L ..... .....
			L ..... 20000.dat
    L ..... grid_B
	    L ..... mean.dat
		L ..... time_series
			L ..... 10000.dat
			L ..... 10020.dat
			L ..... .....
			L ..... 20000.dat
.....
.....
L ..... a1-1
    L ..... README.md
    L ..... grid_A
	    L ..... mean.dat
		L ..... time_series
			L ..... 10000.dat
			L ..... 10020.dat
			L ..... .....
			L ..... 20000.dat
    L ..... grid_B
	    L ..... mean.dat
		L ..... time_series
			L ..... 10000.dat
			L ..... 10020.dat
			L ..... .....
			L ..... 20000.dat
.....
.....
L ..... f5-1
    L ..... README.md
    L ..... grid_A
	    L ..... mean.dat
		L ..... prediction.dat
		L ..... time_series
			L ..... 10000.dat
			L ..... 10020.dat
			L ..... .....
			L ..... 20000.dat
    L ..... grid_B
	    L ..... mean.dat
		L ..... prediction.dat
		L ..... time_series
			L ..... 10000.dat
			L ..... 10020.dat
			L ..... .....
			L ..... 20000.dat
.....
.....
L ..... j5-3
    L ..... README.md
    L ..... grid_A
	    L ..... mean.dat
		L ..... time_series
			L ..... 10000.dat
			L ..... 10020.dat
			L ..... .....
			L ..... 20000.dat
    L ..... grid_B
	    L ..... mean.dat
		L ..... time_series
			L ..... 10000.dat
			L ..... 10020.dat
			L ..... .....
			L ..... 20000.dat