Artificial Intelligence Machine Learning in Marine Hydrodynamics
Name
v009t13a028-omae2018-77599.pdf
Size
1.65 MB
Format
Adobe PDF
Checksum (MD5)
8855799c524579fb58fcfd9153cb1994
Author(s)
Sclavounos, Paul D
Ma, Yu
Date Issued
June 2018
Journal
Proceedings of the ASME 2018 37th International Conference on Ocean, Offshore and Arctic Engineering
Publisher
American Society of Mechanical Engineers
Citation
Sclavounos, Paul D., and Yu Ma. “Artificial Intelligence Machine Learning in Marine Hydrodynamics.” Proceedings of the ASME 2018 37th International
Conference on Ocean, Offshore and Arctic Engineering,17-22 June, Madrid, Spain, ASME, 2018. © 2018 ASME
Version
Final published version
Abstract
Artificial Intelligence (AI) Support Vector Machine (SVM) learning algorithms have enjoyed rapid growth in recent years with applications in a wide range of disciplines often with impressive results. The present paper introduces this machine learning technology to the field of marine hydrodynamics for the study of complex potential and viscous flow problems. Examples considered include the forecasting of the seastate elevations and vessel responses using their past time records as "explanatory variables" or "features" and the development of a nonlinear model for the roll restoring, added moment of inertia and viscous damping using the vessel response kinematics from free decay tests as "features". A key innovation of AI-SVM kernel algorithms is that the nonlinear dependence of the dependent variable on the "features" is embedded into the SVM kernel and its selection plays a key role in the performance of the algorithms. The kernel selection is discussed and its relation to the physics of the marine hydrodynamic flows considered in the present paper is addressed.
MIT Department
Massachusetts Institute of Technology. Department of Mechanical Engineering
Terms of Use
Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use.
Persistent DSpace Link
DOI of Published Version
https://doi.org/10.1115/OMAE2018-77599
