Machine Learning for Vortex Induced Vibration in Turbulent Flow

2021/03/10 by Xiaodong Bai, Wei Zhang, Bai, Xiaodong +1 · 1 citation
Engineering · Mathematics · Physics and Astronomy · #Classical mechanics #Compressibility #Computational fluid dynamics #FOS: Physical sciences #Flow (mathematics) #Fluid Dynamics (physics.flu-dyn) #Fluid Dynamics and Turbulent Flows #Fluid Dynamics and Vibration Analysis #Laminar flow #Mathematics #Mechanics #Model Reduction and Neural Networks #Navier–Stokes equations #Physics #Reynolds number #Reynolds-averaged Navier–Stokes equations #Turbulence #Vortex #Vortex shedding #Vortex-induced vibration #physics.flu-dyn

paper · pdf · doi:10.48550/arxiv.2103.05818

published in arXiv (Cornell University) (Cornell University) · 22 pages, 17 figures

arxiv created 2021/03/10 · openalex publication_date 2021/03/10 · arxiv updated 2021/03/11 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Vortex induced vibration (VIV) occurs when vortex shedding frequency falls close to the natural frequency of a structure. Investigation on VIV is of great value in disaster mitigation, energy extraction and other applications. Following recent development in machine learning on VIV in laminar flow, this study extends it to the turbulent region by employing the state-of-the-art parameterised Navier-Stokes equations based physics informed neural network (PNS-PINN). Turbulent flow with Reynolds number Re = 10⁴, passing through a cylinder undergoing VIV motion, was considered as an example. Within the PNS-PINN, an artificial viscosity ν_t is introduced in the Navier-Stokes equations and treated as a hidden output variable. A recently developed Navier-Stokes equations based PINN, termed NSFnets (Jin et al., J COMPUT PHYS 426: 109951, 2021), was also considered for comparison. Training datasets of scattered velocity and dye trace concentration snapshots, from computational fluid dynamics (CFD) simulations, were prepared for both the PNS-PINN and NSFnets. Results show that, compared with the NSFnets, the PNS-PINN is more effective in inferring and reconstructing turbulent flow under VIV circumstances. The PNS-PINN also shows the capability to deal with unsteady and multi-scale flows in VIV. Inspired by the Reynolds-Averaged Navier-Stokes (RANS) formulation, by implanting an additional artificial viscosity, the PNS-PINN is free of complex turbulence model closure, thus may be applied for more general and complex turbulent flows.

Citations

• Two-equation eddy-viscosity turbulence models for engineering applications
• Vortex-induced vibrations of a circular cylinder at low Reynolds numbers
• VORTEX-INDUCED VIBRATIONS
• Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations
• Understanding and mitigating gradient pathologies in physics-informed neural networks
• Hidden fluid mechanics: Learning velocity and pressure fields from flow visualizations
• NSFnets (Navier-Stokes Flow nets): Physics-informed neural networks for the incompressible Navier-Stokes equations

Related

• Deep learning based on PINN for solving 2 D0F vortex induced vibration of cylinder with high Reynolds number
• A transfer learning enhanced the physics-informed neural network model for vortex-induced vibration
• VIVALDy: A Hybrid Generative Reduced-Order Model for Turbulent Flows, Applied to Vortex-Induced Vibrations
• Physics-enhanced Neural Operator for Simulating Turbulent Transport
• Predictive Limitations of Physics-Informed Neural Networks in Vortex Shedding