The turbulence research group at McMaster University has developed a pseudo spectral code for calculating turbulent flow through periodic cylinder arrays in three dimensions.  The principal application of their code is to investigate the fluid elastic instability of heat exchanger tube bundles in steam generators (e.g. in nuclear reactors).

    The coupling of hydrodynamic drag and lift forces with the mechanical vibration of the tubes can produce extreme resonance's leading to catastrophic failure of the heat exchanger.  It is therefore important to be able to predict which configurations lead to stable operation.

    The main goal of this project is to parallelize this code (in part using available parallelizing FFT's), and then evaluate its parallel performance on Sharcnet McMaster's 128  node cluster Idra .  We studied the code structure, proposed a parallelization scheme, and evaluated its efficiency. In particular, we investigated the parallel performance of the code and the properties of the flow as three-dimensionality is increased by adding more degrees of freedom in the direction of the cylinder axis.  This study should answer the following open scientific question: engineering experiments suggest that the flow remains two-dimensional (even at high Reynolds number), but it is unclear physically how this is possible given the inherent three-dimensional instability of turbulent flows at moderate Reynolds numbers.  The numerical simulation will allow us to determine which quantities (e.g. fluid forces) do not change as the flow is allowed to become more three-dimensional, and how this two-dimensionality is maintained.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Introduction

Parallelization