Date of Award:

2013

Document Type:

Thesis

Degree Name:

Master of Science (MS)

Department:

Mechanical and Aerospace Engineering

Advisor/Chair:

Robert E. Spall

Abstract

Computational fluid dynamics (CFD) is commonly used to visualize and understand complicated fluid flow and heat transfer in many industries. It is imperative to validate the CFD computer models in order to avoid costly design choices where experimentation cannot be used to ratify the predictions of computer models. Assessments of CFD computer models in the literature conclude that significant errors occur in computer model predictions of fluid flow influenced by buoyancy forces.

The Experimental Fluid Dynamics Laboratory at Utah State University constructed a wind tunnel with which to perform experiments on buoyancy induced fluid flow. The experiments measured the heat transfer and fluid velocity occurring in the buoyant flows to be used to validate computer models. Additional experimental measurements at the inlet and around the walls from each experiment allowed the computer models to simulate the fluid flow with realistic boundary conditions.
For this study, four experiments were performed, including two cases where the buoy- ancy influence was significant, and two where it was not. For each set of two cases, one experiment was performed where the heat transfer occurred from a wall of the wind tunnel held at constant temperature and in the other experiment the wall temperature fluctuated axially.

This study used the experimental data to validate computer models available in the general purpose CFD software STAR-CCM+, including the k − ε models: realizable two- layer, standard two-layer, standard low-Re, v2 − f, the k − ω models from Wilcox and Menter, and the Reynolds stress transport and Spalart–Allmaras models. The k − ε stan- dard low-Re model was found most capable overall of predicting the fluid flow and heat transfer that occurred in the flows where the buoyancy influence was significant. For the experimental cases where the buoyancy influence was less significant, the validation results were inconsistent.