Sustainable Marine Structures

Active Learning-Based Construction of a Hydrodynamics Database for Critical Reynolds Numbers and Its Application to Vortex-Induced Vibrations Predictions

Hao Liu

Department of Engineering Mechanics, School of Ocean and Civil Engineering, Shanghai Jiao Tong University, Shanghai 200240, China Ministry of Education (MOE) Key Laboratory of Hydrodynamics, Shanghai Jiao Tong University, Shanghai 200240, China

Jiasong Wang

Shanghai Jiao Tong University

Yigang Gong

Department of Engineering Mechanics, School of Ocean and Civil Engineering, Shanghai Jiao Tong University, Shanghai 200240, China Ministry of Education (MOE) Key Laboratory of Hydrodynamics, Shanghai Jiao Tong University, Shanghai 200240, China

Zhongxu Tian

College of Engineering Science and Technology, Shanghai Ocean University, Shanghai 201306, China

Yuanjiang Chang

Centre for Offshore Engineering and Safety Technology, China University of Petroleum (East China), Qingdao 266580, China

DOI: https://doi.org/10.36956/sms.v8i2.3258

Received: 16 April 2026; Published: 17 June 2026

Abstract

For the cross-flow vortex-induced vibration (VIV) response of marine risers, comprehensive research has been conducted within the subcritical Reynolds number range. Despite ongoing efforts toward full-scale marine riser deployment, the understanding of VIV dynamics in the critical Reynolds regime remains limited. This is mainly due to the scarcity of high-fidelity data capturing key flow phenomena in this regime, including laminar separation bubbles, turbulent reattachment, and the drag crisis. This study established a hydrodynamic database for the critical Reynolds number regime using a custom high-resolution algorithm platform for incompressible flow (HRAPIF), leveraging Bayesian deep active learning (BDAL) to drive computational fluid dynamics (CFD) simulations. BDAL efficiently sampled the target in parameter space and established hydrodynamics coefficients with fewer large eddy simulation (LES) runs. This database was embedded in the in-house code DAVIV for rapid prediction of VIV. Validated against Chaplin's experimental data, the prediction accuracy of DAVIV is comparable to that of SHEAR7 at subcritical Reynolds numbers. For full-scale risers, DAVIV provides more balanced performance and avoids excessive amplitude prediction at the critical Reynolds number. The result from DAVIV has good consistency with high-fidelity CFD. Overall, this framework addresses the bottleneck in VIV prediction accuracy within the critical Reynolds number regime, and provides a practical and low-cost tool for marine engineering.

Keywords: Vortex Induced Vibration, Critical Reynolds Numbers, Computational Fluid Dynamics, Riser Response Prediction

References