Roll damping predictions using physics-based machine learning
Roll damping predictions using physics-based machine learning
Computational Fluid Dynamics simulations and Machine Learning models are useful predictions tools that have the potential to work even better when used together. This paper presents a physics- based machine learning approach that supplements standard regression basis functions, such as polynomials, with simple physical models of the system. This mitigates the data dependence of machine learning predictions, and the associated computational cost of generating the training set simulations. We illustrate this method by increasing the accuracy of roll-damping power coefficient predictions by 50 to 200% using O(10) training examples.
Weymouth, Gabriel
b0c85fda-dfed-44da-8cc4-9e0cc88e2ca0
March 2019
Weymouth, Gabriel
b0c85fda-dfed-44da-8cc4-9e0cc88e2ca0
Weymouth, Gabriel
(2019)
Roll damping predictions using physics-based machine learning.
Computer Applications and Information Technology in the Maritime Industries, , Tullamore, Ireland.
25 - 27 Mar 2019.
8 pp
.
Record type:
Conference or Workshop Item
(Paper)
Abstract
Computational Fluid Dynamics simulations and Machine Learning models are useful predictions tools that have the potential to work even better when used together. This paper presents a physics- based machine learning approach that supplements standard regression basis functions, such as polynomials, with simple physical models of the system. This mitigates the data dependence of machine learning predictions, and the associated computational cost of generating the training set simulations. We illustrate this method by increasing the accuracy of roll-damping power coefficient predictions by 50 to 200% using O(10) training examples.
Text
Weymouth 2019 COMPIT
- Author's Original
More information
Accepted/In Press date: 1 February 2019
Published date: March 2019
Venue - Dates:
Computer Applications and Information Technology in the Maritime Industries, , Tullamore, Ireland, 2019-03-25 - 2019-03-27
Identifiers
Local EPrints ID: 429892
URI: http://eprints.soton.ac.uk/id/eprint/429892
PURE UUID: 73b06541-4762-4f95-8070-f353410c0920
Catalogue record
Date deposited: 08 Apr 2019 16:30
Last modified: 16 Mar 2024 04:15
Export record
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics
