Dynamical modeling of vortex-induced vibration using frequency-constrained autoencoder neural networks

In most flows around a structure or object, the flow behind the object is characterized by a region of unsteady reverse flow. This unsteady fluid motion in the object wake induces a force upon the object or structure, which can cause flexible structures to vibrate. An ongoing engineering challenge is to understand how the wake and structure interact, and how to minimize these vibrations. This project proposes a novel method for developing simplified models, also known as Reduced Order Models, which capture the underlying mechanisms of these interactions using a combination of statistical and neural network methods. These methods find a simplified representation for the velocity in the structure wake, and then find equations that model how it changes in time. These equations can be used to understand the physics of the interactions, and to design better methods to reduce the vibrations.

Faculty Supervisor:

Chris Morton;Robert Martinuzzi

Student:

Partner:

Technische Universität Berlin

Discipline:

Engineering

Sector:

Aerospace; Artificial Intelligence; Green/Alternative Energy

University:

University of Calgary

Program: