| Link | https://publications.isope.org/proceedings/ISOPE/ISOPE%202024/data/pdfs/459-2024-TPC-0476.pdf |
|---|
| Author | Search for: He, Moqin1; Search for: Akinturk, Ayhan1; Search for: Seo, Dong Cheol1ORCID identifier: https://orcid.org/0000-0002-5818-7475; Search for: Mak, Lawrence1; Search for: Zaman, Hasanat1 |
|---|
| Affiliation | - National Research Council of Canada. Ocean, Coastal and River Engineering
|
|---|
| Format | Text, Article |
|---|
| Conference | The 34th International Ocean and Polar Engineering Conference June 16–21, 2024 Rhodes, Greece |
|---|
| Subject | artificial intelligence; implementation; application; prediction accuracy; dataset; implementation; machine learning engineering application; machine learning; Gaussian process regression; dimensional analysis; regression model; engineering application |
|---|
| Abstract | Many machine learning applications in engineering fields have been conducted without enough description on implementation details during dataset preparations. Theoretically a pure data driven machine learning procedure works well when a single measuring unit, or a combination of base units, is contained in the dataset. Datasets in engineering fields usually contain multiple measuring units. In this case, both unit and quantity contained in the data are important to model data correlations. This paper presents steps of dataset preparation including dimensional analysis and variable transformations of inputs and responses to bring physical information out from units into machine learning. Through simple case studies this paper also demonstrates steps to train a model with data from multiple objects as applicable, called a collective model. The collective model is trained using the extended dataset constructed by merging data from all included objects as well as some parameters identifying the individual objects used. Results of this study indicate that predictions by the individually trained models for the individual objects generally are more accurate than those obtained from the collective model. It has also been observed that incorporating dimensional analysis can improve the prediction accuracy of the collective model. |
|---|
| Publication date | 2024-06-16 |
|---|
| Publisher | International Society of Offshore and Polar Engineers |
|---|
| In | |
|---|
| Language | English |
|---|
| Peer reviewed | Yes |
|---|
| Export citation | Export as RIS |
|---|
| Report a correction | Report a correction (opens in a new tab) |
|---|
| Record identifier | 84c8d61c-b47f-4461-bc49-c3b2357f645d |
|---|
| Record created | 2025-06-19 |
|---|
| Record modified | 2025-09-12 |
|---|
