No abstract

Turbulent flow in Z-shape duct configuration is investigated using Reynolds stress model (RSM) and ζ-f model and compared to experimental results. Both RSM and ζ-f models are based on steady-state RANS solutions. The focus was on regions where the RSM has over- or underpredicted the flow when compared to the experimental results and on regions where there are flow separations and high turbulence. The performance of predicting the flow reattachment length in each model is studied as well. RSM has shown the mean flow velocity profile results match reasonably well with the experiment. Advanced ζ-f turbulence model is introduced as user-defined function (UDF) code and applied to the Z-shape duct. It is found that the turbulent kinetic energy production in ζ equation is much easier to reproduce accurately. Both mean velocity gradient and local turbulent stress terms are also much easier to be resolved properly. The current research has found that not only ζ-f model takes less time to complete the simulation but also the mean flow velocity profile results are in better agreement with the experimental data than the RSM although both are coupled steady-state RANS. ζ-f model numerically resolved both the flow separation and reattachment regions better than the RSM. The current numerical results from ζ-f model are attractive and encouraging for wall-bounded flow applications where flow separation and flow reattachment are important for the flow mechanism.

A numerical study is conducted of the transient flow in a z-shape threedimensional duct using large-eddy simulation (LES) with dynamic eddy viscosity subgrid-scale (SGS) model with a fully structured grid system. The numerical results of the velocity profiles are quantitatively validated against experimental data for zshape ducts with various lateral separation distance configurations. The framework of the current LES model has been studied and discussed and the performance of LES in predicting the flow in z-shape ducts as a function of separation distances was evaluated. LES predictions of the mean flow velocity profiles are in good agreement (within the experimental uncertainty) with experimental data for the investigated wide range of l/d configurations. This is attributed to the well-resolved large-scale flow structures. Some slight over-predictions and under-predictions were found at certain separation distances. These numerical errors are due to the limited modeling approach to predict small eddies structures with the current SGS model. The main key features of the flow after the first elbow are also identified as separation and re-attachment regions. Some discrepancies are identified for lateral separation distances at sections x/d = 3 and x/d = 5 inside the flow transition regions. These discrepancies are believed to be inherited from the upstream flow numerical errors that arise in the non-uniform flow mixing regions. The potential remedy includes applying finer mesh resolution and/or higher order spatial discretization to accurately resolve the local velocity gradients and the complex flow structures. ABOUT THE AUTHOR Mohammed Karbon is currently a PhD candidate in the Department of Mechanical and Industrial Engineering at Qatar University. Karbon expertise is in Computational Fluid Dynamics and Turbulence Modeling working under the supervision of Professor Ahmad K. Sleiti. He has many years of industrial work experience before joining the group of Prof. Ahmad. Karbon is also familiar with several computational skills and commercial software packages including ANSYS FLUENT and others.

No abstract