Session: 16-01-01: Poster Session: NSF-Funded Research (Grad & Undergrad)

Paper Number: 99381

99381 - Numerical Wave Experiments: Physics-Based and Data-Driven Approaches 

Freak waves, which are also known as extreme waves or rogue waves, have posed considerable
challenges for maritime traffic and offshore operations. The crest heights of these waves exceed
the significant wave height by a factor of 1.25. In the ocean, these wave tend to appear suddenly
on the surface and their occurrence is much more frequent than has been thought previously. The
devastating impact of these extreme events has drawn significant efforts from across the globe to
better understand them. Several mechanisms have been proposed for the formation of these
waves. Modulational instability, which is also better known as Benjamin-Feir instability, is an
instability mechanism that has been proposed to explain such energy localizations by using
nonlinear mechanics. To explore such mechanisms as well as other mechanisms such as
focusing, numerical and/or physical wave tank experiments can be beneficial. Given the
computing power available today and the parameter space to be explored, numerical wave tank
experiments can provide an appealing avenue. Here, a combination of physics-based and data-
driven approaches are pursued to conduct these numerical experiments.
Over the last several decades, Lagrangian particle methods have been employed successfully for
the investigation of hydrodynamic and astrophysical flows. Building on such efforts, numerical
investigations have been carried out here by using Weakly Compressible Smoothed Particle
Hydrodynamics (WCSPH) to examine modulational instabilities in plane perturbed waves.
General Process Graphics Processing Unit (GPGPU) computing has been utilized to accelerate
the simulation process. The numerical scheme, which has been validated by using regular and
irregular wave propagation scenarios, has been used to carry out parametric studies on
unidirectional propagation of perturbed regular waves in a numerical wave tank. Specifically,
surface elevation histories have been used to study associated eigenvalue and frequency spectra.
Furthermore, the effect of depth has also been investigated in the context of such perturbed
waves.
As one avenue towards leveraging data-driven approaches in complement with physics-based
simulations, recourse has been taken to developing an efficient prediction tool based on a deep
leaning framework. Here,. deep learning frameworks have been utilized for the prediction of
missing wave data in wave records obtained through measurements, say, from buoys in the
ocean. The effectiveness of the different types of neural networks have been examined by using
different available wave data to determine an appropriate model for these predictions. In
concert, with physics-based approaches, these data-driven approaches can provide a powerful
numerical tool for conducting wave experiments to study waves and energy localizations in the
ocean.

Presenting Author: Samarpan Chakraborty University of Maryland, College Park

Presenting Author Biography: I am a PhD student working in the field of non-linear dynamics mainly focusing on the investigations of the mechanisms leading to extreme oceanic events and their subsequent prediction.

Authors:

Samarpan Chakraborty University of Maryland, College Park
Balakumar Balachandran University of Maryland
Kayo Ide University of Maryland, College Park