Machine Learning for Turbulence Modelling

Machine learning has revolutionized a variety of fields in the past decade, due to increasing availability of data and processing power. Engineering simulations of most industrially relevant fluid flows (e.g aircraft design and turbomachinery) require modelling of turbulent fluctuations in the flow. For the turbulence modelling community, which has seen widespread stagnation, machine learning offers a clear path to improve model accuracy, estimate uncertainty, and develop new data driven models. While the potential for machine learning in the field of turbulence modelling is clear, the number of thorough investigations is limited. This project involves a detailed investigation into improvements and testing of a neural network architecture for turbulence modelling, and implementation of this method in a practical engineering simulation setting. The project represents a critical leap forward in revolutionizing the field of turbulence modelling augmentation by machine learning.

Faculty Supervisor:

Fue-Sang Lien

Student:

Partner:

University of Manchester

Discipline:

Engineering

Sector:

Education

University:

University of Waterloo

Program: