With the current development of the computer industry, CFD simulations have become the widespread standard in the industry, forming a baseline tool for numerous designs and safety procedures. This extensive dependence on the CFD codes rather than experiments raises the issue of the reliability of the results obtained from these codes. This thesis is intended to study the dependence of the CFD results on the grid types, numerical schemes and turbulence models. Additionally, comparisons between a general purpose commercial code STAR-CCM+ and a specialized code FDS are presented towards the end of this thesis. To study the numerical errors introduced by the grids and schemes, a laminar flow induced by natural convection inside a square cavity was considered first. Using Richardson’s extrapolation, a grid independent solution was calculated and compared with the results obtained from different grid types and schemes for Rayleigh (Ra) numbers , and . Comparison plots showed a higher dependence of the accuracy of the results on the cell shapes along with the order of the scheme and the cell size. Additionally, with the same cavity a grid dependence study for the and model has been done at .To test the reliability of the Quasi-DNS performed by an Unstructured Finite Volume (FV) CFD code, Turbulent Kinetic Energy (TKE) budgets should be calculated. User subroutines were developed to calculate the budgets of the TKE and to verify the user subroutines, prior to coaxial cylinder test case, a Q-DNS of the channel flow at has been performed using different grid configurations and numerical schemes. Results obtained from the Q-DNS of the channel flow on the polyhedral cells with the bounded central differencing scheme were found to be in good agreement with the reference DNS data. After the validation test case, a Q-DNS of the buoyancy driven turbulent flow inside a horizontal annular cavity at a high Rayleigh number, Ra = 1.18x109 with outer to inner cylinder ratio of 4.85 was carried out using a commercial code. Comparisons of Q-DNS results with low-Re URANS models, and model, showed that the latter models are able to capture the general flow features but fail to predict the large unsteadiness and high turbulence levels in the plume. However, local heat transfer rates along the inner and outer cylinder walls are on average of acceptable accuracy for engineering purposes. Finally, a full scale industrial test case of a fire in a compartment has been simulated. Both URANS ( model) and LES (Smagorinsky model) approaches are applied to model the turbulence with and without incorporating the combustion modelling. A comparison of the CFD results with the experimental data showed that for building fire simulations, accuracy of the results is more sensitive to the correlations used in the combustion modelling rather than the type of the turbulence model.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:607039 |
| Date | January 2013 |
| Creators | Zaidi, Imama |
| Contributors | Laurence, Dominique |
| Publisher | University of Manchester |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | https://www.research.manchester.ac.uk/portal/en/theses/reliability-of-cfd-for-buoyancy-driven-flows-in-industrial-applications(f053b55d-7514-479a-a905-58a2a048aaee).html |
Page generated in 0.0019 seconds