Date: 12 December, 2024

Dataset Title: Machine Learning Dataset to Support Paper "A comparison of machine learning classifiers in predicting safety for a multi-component dynamic system representation of an autonomous vessel"

Dataset Contact: Matthew Collette, mdcoll@umich.edu

Dataset Creators: 
Name: Matthew Collette
Email: mdcoll@umich.edu
College of Engineering, University of Michigan, Ann Arbor, MI 
ORCID: https://orcid.org/0000-0002-8380-675X

Name: Brendan Sulkowski
Email: bsulkow@umich.edu 
College of Engineering, University of Michigan, Ann Arbor, MI 
ORCID: https://orcid.org/0009-0000-9970-3918

Description:
This data set contains: 1) Globel locations and 0h, 24h, 72h, 144h, and 240h sea-state forecasts for 1000 select locations in May 22 2023 captured from outputs in the European Centre for Medium-Range Weather Forecasts(ECMWF) Creative Commons Licensed Products https://www.ecmwf.int/en/forecasts 2) Spring-mass-damper system parameters from random sampling, and various approximations of these parameters 3) Pass-fail motion threshold data that was used to train different ML models. 

Use and Access:
This data set is made available under a Creative Commons Public Domain license (CC0 1.0).

To Cite and Understand Methods Behind Data:
Brendan Sulkowski, Matthew Collette, "A comparison of machine learning classifiers in predicting safety for a multi-component dynamic system representation of an autonomous vessel", Applied Ocean Research 2024 Open Access - https://doi.org/10.1016/j.apor.2024.104368

File Inventory:
MLPaperData.csv - dataset used to train models in the paper. 

Data Description:
The data is contained in a CSV file. Each row represents one data point with one simulated spring-mass-damper and actual weather for a specific location on the globe, with the columns as attributes. Column descriptions the first row in the file, and are summarized here by column number in the file: 
 
 (1) mass value for spring-mass-damper (SMD), kg
 (2) spring stiffness value for SMD, N/m 
 (3) damping value for SMD, N*s/m
 (4-5) Significant wave height (meters) and period, Tp, (seconds), at 0000 on 22 May 2023 (hindcast) 
 (6-8) Category values for stiffness from simulated inspection (see paper for complete description of binning procedure 0= failed, 1=worn, 2=health) 
 (9-17) noise-contaminated estimates of m, k, and c (as labeled) see paper for complete details of noise parameters. Same units a (1-3)
 (18-19) Lat/Long of each sample location on the globe 
 (20-23) 24, 72, 144, and 240 prior forecasts for Significant Wave Height at 0000 on 22 May 2023 (archived forecasts, meters) 
 (23-27) 24, 72, 144, and 240 prior forecasts for Tp at 0000 on 22 May 2023 (archived forecasts, seconds) 
 (28) Pass (0 - stayed below ) or fail (1 - stayed above) of motion with true m/c/k/ and hindcast Hs/Tp failed motion criteria.  Criteria was if average of 1/1000th highest responses exceeded 5m of motion.  

 Funding:
 Office of Naval Research Code 331 N00014-21-1-2795.