Flow past a square prism

Flow past a square prism: A numerical study

Nanda, S.

Terwisga, T.V. (mentor)

Mechanical, Maritime and Materials Engineering

Maritime Technology

2016-05-24

The ERCOFTAC test case for flow past a square-prism at a Reynolds number of 21,400 was studied numerically using the flow solver ReFRESCO. The main objectives were to evaluate the performance of a Reynolds Averaged Navier-Stokes (RANS) based turbulence models in Unsteady RANS (URANS) simulations, identify sensitivities to numerical parameters and find out if Detached-Eddy Simulation (DES) based turbulence models provide lower modelling errors than in URANS simulations for this practical test-case. In this respect, the effect of the iterative error and the discretization error using k-omega SST-2003 model [15] was studied and turbulence models such as the SST-DDES turbulence model [19] and the TNT-XLES turbulence model [20] were also applied. The results using the URANS results have highlighted the importance of small temporal and spatial resolution in the calculation domain in order to reduce the effect of the discretization errors. Moreover, large time-integration periods (larger than 60 shedding cycles) should be required to ensure that the statistical noise does not affect the solution Verification procedure which would otherwise hamper the estimated numerical uncertainties. As expected from the studies of Pereira et al. [7], the results with the DES based turbulence models have shown that numerical settings for URANS are insufficient for DES and large comparison errors have been found for many quantities such as the time-averaged pressure distribution at the surface and the integral quantities. The numerical studies with the k-omega SST model [15] have shown that if the numerical uncertainties identified in the study are controlled, Verification & Validation studies using this model could provide low modelling errors for this test-case for the investigated mean flow quantities. For the DES results, since the poor comparisons with experimental results with respect to URANS results using the k-omega SST (2003) turbulence model are caused by poor numerical settings, studies using finer grids, smaller time-steps and lower iterative criteria than the ones used in the present work are recommended.

square prism
Detached Eddy Simulation
URANS
turbulence modelling

http://resolver.tudelft.nl/uuid:ba8daa3d-8fb8-4900-8e44-4f6b7c2f8754

Student theses

master thesis

(c) 2016 Nanda, S.