Date of Award:


Document Type:


Degree Name:

Doctor of Philosophy (PhD)


Mechanical and Aerospace Engineering

Committee Chair(s)

Barton L. Smith


Barton L. Smith


Robert Spall


Heng Ban


Aaron Katz


Blake Tullis


A CFD validation study of unsteady flow through a confined bank of cylinders is performed to demonstrate validation of modeling and simulation (M&S) for time-varying quantities. Additionally, this dissertation assesses appropriate unsteady system response quantities (SRQ’s) and proper estimates of the numerical and experimental uncertainty. CFD simulations include the Unsteady Reynolds-Averaged Navier-Stokes (URANS) k-ω model and two variations of the Detached Eddy Simulation (DES) model. Data is acquired using a non-intrusive measurement technique, called particle image velocimetry (PIV), coupled with time-varying pressure measurements along the facility walls. These data are used to validate the numerical results. The numerical simulations are designed to closely resemble the experimental facility boundary and inlet conditions and be geometrically identical.

Requirements for validation studies include 1) non-invasive measurement techniques, such as PIV, and 2) accurate quantification of the measurement uncertainty. However, accurate - and even adequate - local, instantaneous PIV uncertainty quantification is often difficult and ambiguous, particularly for higher-order terms (e.g. temporal variations and turbulence fluctuations). Thus, the effects of four PIV error sources on velocity and turbulence statistics are assessed and demonstrated within this study: particle image size, particle image displacement, particle image density, and gradients within the flow. Results of this uncertainty study are found to accurately estimate the uncertainty from these error sources and are extended to generate PIV uncertainty estimates for the validation study.

Unsteady spatial and temporal validation SRQ’s, such as frequency, fluctuations, autocorrelations, and correlations, are assessed to determine their effectiveness as validation quantities in the bank of cylinders experiment. Some of the models accurately predicted frequencies present in the velocity, bulk velocity, and pressure SRQ’s for cylinders 1, 4, and 5. However, additional, non-physical frequencies are predicted for the remaining cylinders from all models. As expected, the temporal behavior of the DES was generally far superior to that of the URANS model. While some models are capable of predicting fluctuation amplitudes, and autocorrelation and correlation time-scales throughout the facility for most SRQ’s, this is generally not observed for the remaining models. A grid convergence study shows typical global quantities (such as pressure losses) converge well while temporal quantities converge poorly for the same grids.




This work made publicly available electronically on May 9, 2012.

Included in

Philosophy Commons