Doctoral Defense in Computional Engineering - Seyedreza Hassanianmoaref

Doctoral candidate:
Seyedreza Hassanianmoaref

Title of thesis:
Design and Evaluation of Parallel & Scalable Machine Learning Approaches in Computational Fluid Dynamics Applications

Opponents:
Dr. Andrea Beck, Professor of Numerical Methods in Fluid Mechanics at the Institute of Aerodynamics and Gas Dynamics and the Stuttgart Center for Simulation Science at the University of Stuttgart, Germany, Dr. Amir Hossein Shiravi, Professor of Mechanical Engineering Department at the Jundi Shapur University of Technology, Iran

Advisor:
Dr. Morris Riedel, Professor at the Faculty of Industrial Engineering, Mechanical Engineering and Computer Science at the University of Iceland and Head of High Productivity Data Processing research group at the Juelich Supercomputing Centre, Germany

Also in the doctoral committee:
Dr. Ásdís Helgadóttir, Associate Professor of Industrial Engineering, Mechanical Engineering and Computer Science, University of Iceland Dr. Pedro Simões Costa, Assistant Professor of Faculty of Mechanical Engineering TU Delft, Netherland

Chair of Ceremony:
Dr. Rúnar Unnþórsson, Professor and Head of Faculty of Industrial Engineering, Mechanical Engineering and Computer Science, University of Iceland

Abstract:
Turbulent flow, an unsolved problem in physics, manifests in numerous industrial and natural contexts. Computational fluid dynamics (CFD) and experimental fluid dynamics (EFD) are primary methods to study turbulent flow, yet both face limitations due to theoretical understanding, scale, and cost. The random and nonlinear nature of turbulent flow further complicates these approaches. Recently, deep learning has shown significant promise in predicting nonlinear phenomena. This thesis proposes and develops a novel approach to link Lagrangian particle tracking (LPT) in fluid dynamics with sequential deep learning (DL) models to predict turbulent flow. High-performance computing (HPC) is essential for these DL models, and this research leverages cutting-edge HPC systems provided by EuroCC 1 and 2. The outcomes of this thesis are also reported in the CoE-RAISE project. Using experimental data from prior PhD student research, the DL models—specifically long short-term memory (LSTM), gated recurrent unit (GRU), and Transformer—demonstrated remarkable success in predicting turbulent flow. Additionally, a wind energy use case was developed to apply the thesis model to an engineering context. The turbulence research community has recognized the thesis's contributions, leading to publications in several respected scientific journals and presentations at numerous conferences.

About the candidate:
Seyedreza Reza was born in Sary, Iran, in 1985. He earned his B.Sc. in Mechanical Engineering from Chamran University in Ahwaz, Iran, in 2009. He then worked in research and development, focusing on fluid dynamics in the energy sector. In 2020, Reza completed his M.Sc. in Sustainable Energy Engineering at Reykjavik University and immediately pursued a Ph.D. in Computational Engineering at the University of Iceland. His primary research focuses on computational fluid dynamics (CFD) applying deep learning (DL) models to analyze and understand turbulence flow problems. Additionally, he has worked on leveraging High-Performance Computing to accelerate, optimize, and scale these models. Reza has two children and enjoys spending his time with them and his wife.