| With the rapid proliferation of hard disk drives into non-traditional applications new demands are placed on the size, speed and reliability of these drives. There has been a strong demand for higher areal density, faster data transfer rates and better reliability. Higher track densities require in the reduction of the available area to position the read-write head, thereby reducing the allowable tolerance for track-misregistration (TMR). On the other hand, higher data transfer rates require higher speeds of disk rotation, which in turn increase the Reynolds number of the air flow and hence the turbulent excitation of the flow.; Numerical simulations of the turbulent flow of air inside model hard disk drives are reported here using a commercial CFD software, CFD-ACE. Even with current supercomputer resources direct numerical simulation (DNS) is not feasible. On the other hand, Reynolds Averaged Navier Stokes (RANS) methods would not capture the essential unsteadiness of the flow. For this reason, large eddy simulation (LES) is the most reliable and accurate method for simulating such flows at reasonable cost. Different flow quantities (velocities, pressure, vorticity) are analyzed, global quantities such as drag on the arm and windage are reported, and the coupled flow structure interaction problem is solved. The pressure and shear stress coupling is done in only one direction, from the flow to the structure. It is observed that rapid vorticity shedding occurs from the sharp corners of the arm, and subsequently this vorticity organizes into turbulent eddies. The highly unsteady wake is transported by the rotating disks and is dissipated along the azimuthal span of the drive. The off-track vibrations of the are approximately 4-5 nanometers in RMS, mainly due to vibrations in the first sway and torsion modes.; After reporting the basic flow features, attention is devoted to the accuracy and validity of the results. Firstly, the behavior and accuracy of subgrid scale models (SGS), which form the central core of the LES technique, are investigated. Three different SGS models are compared with a direct numerical simulation. It is shown that the algebraic dynamic model is the optimal choice for the SGS model. Next, three commercial CFD codes CFD ACE, Fluent and CFX are benchmarked in their ability to solve a standard test problem for LES---the flow across a square cylinder. It is observed that Fluent and CFD-ACE provide accurate results when using the dynamic model, but the results deteriorate when using the Smagorinsky model. CFX displayed the largest deviations in results among the three codes when compared to the experimental data.; To build more credibility into the results, extensive experimental validation is carried out. Validation is carried out against both hot-wire anemometry data (Gross 2003) and particle image velocimetry (Barbier 2006) data. In the context of experimental validation comprehensive grid convergence studies are also performed. It is shown that grids in the 2-2.5 Million cell range are in asymptotic range for convergence. The rates of convergence agree well with the theoretical rates for the discretization schemes. The grid based uncertainty of our results is then estimated to be approximately between 15-30%. Further, the numerical dissipation associated with our grid and temporal numerical scheme is approximated using a simple technique. The artificial dissipation is approximately 18% of the total dissipative processes.; The LES results are then applied to two problems: computation of the flow induced vibrations of the rotating disk and computation of the vibrations of the arm in the presence of flow mitigation devices. In the former, a self developed spectral finite-difference code is used to solve for the elastic vibrations of the rotating disk. In the latter, HDD casings which claim reduction in flow induced vibrations by the use of small geometrical modifications are investigated. These modifications i... |
