Browse by author
Lookup NU author(s): Zhu Huang, Runzi Zhao, Kai Wang, Professor Zhiqiang HuORCiD
This work is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0).
© 2026 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group. Wind-assisted propulsion using rigid wing sails is a promising solution for reducing greenhouse gas emissions in shipping, yet accurate full-scale performance prediction remains challenging due to simplified physical models and complex operating conditions. This study employs an AI-based framework built on the Software-in-the-Loop combined Artificial Intelligence method for Dynamic Analysis (SADA) to improve propulsion performance prediction for a rigid wing sail-equipped vessel. A semi-empirical numerical model is coupled with a reinforcement learning module, where key discipline parameters (KDPs) are adaptively calibrated using the Deep Deterministic Policy Gradient (DDPG) algorithm. The framework reduces discrepancies between predicted engine power and engine power estimated from operational measurements while preserving physical interpretability. Validation using steady-state operating data demonstrates that engine power prediction errors are reduced from 10–17% to below 1% compared with the baseline model. The framework is flexible and can be extended to incorporate more sophisticated physics-based models in future studies.
Author(s): Huang Z, Zhao R, Wang K, Hu C, Huang L, Hu Z
Publication type: Article
Publication status: Published
Journal: Ships and Offshore Structures
Year: 2026
Pages: Epub ahead of print
Online publication date: 31/05/2026
Acceptance date: 19/05/2026
Date deposited: 23/06/2026
ISSN (print): 1744-5302
ISSN (electronic): 1754-212X
Publisher: Taylor and Francis Ltd
URL: https://doi.org/10.1080/17445302.2026.2679696
DOI: 10.1080/17445302.2026.2679696
Data Access Statement: The datasets supporting the findings of this study are available in Zenodo at https://doi.org/10.5281/zenodo.18521136
Altmetrics provided by Altmetric
