Font Size: a A A

Direct numerical simulation of shockwave and turbulent boundary layer interactions

Posted on:2008-07-16Degree:Ph.DType:Dissertation
University:Princeton UniversityCandidate:Wu, MinweiFull Text:PDF
GTID:1440390005470532Subject:Engineering
Abstract/Summary:
Direct numerical simulations (DNS) of a shockwave/turbulent boundary layer interaction (STBLI) at Mach number 3 and Reynolds number based on the momentum thickness of 2300 are performed. A 4th-order accurate, bandwidth-optimized weighted-essentially-non-oscillatory (WENO) scheme is used and the method is found to be too dissipative for the STBLI simulation due to the over-adaptation properties of this original WENO scheme. In turn, a relative limiter is introduced to mitigate the problem. Tests on the Shu-Osher problem show that the modified WENO scheme decreases the numerical dissipation significantly. By utilizing a combination of the relative limiter and the absolute limiter described by Jiang & Shu [32], the DNS results are improved further. The DNS data agree well with the reference experiments of Bookey et al. [10] in the size of the separation bubble, the separation and reattachment point, the mean wall-pressure distribution, and the velocity profiles both upstream and downstream of the interaction region.; The DNS data show that velocity profiles change dramatically along the streamwise direction. Downstream of the interaction, the velocity profiles show a characteristic "dip" in the logarithmic region, as shown by the experiments of Smits & Muck [66] at much higher Reynolds number. In the separation region, the velocity profiles are found to resemble those of a laminar flow, yet the flow does not fully relaminarize. The mass-flux turbulence intensity is amplified by a factor of about 5 throughout the interaction, which is consistent with that found in higher Reynolds experiments of Selig et al. [52]. All Reynolds stress components are greatly amplified by the interaction. Assuming that the ow is still two dimensional downstream of the interaction, the components rhou'u', rhov'v', rho w'w', and rho u'w' are amplified by factors of 6, 6, 12, and 24, respectively, where u is the streamwise and w is the wall-normal velocity. However, analyses of the turbulence structure show that the ow is not uniform in the spanwise direction downstream of the interaction. A pair of counter-rotating vortices is observed in streamwise-wall-normal planes in the mean ow downstream of the ramp corner. Taking the three-dimensionality into account, the amplification factors of the Reynolds stresses are greatly decreased. The component rhou'w' is amplified by a factor of about 10, which is comparable to that found in the experiments of Smits & Muck [66]. Strong Reynolds analogy (SRA) relations are also studied using the DNS data. The SRA is found to hold in the incoming boundary layer of the DNS. However, inside and downstream of the interaction region, the SRA relations are not satisfied.; From the DNS analyses, the shock motion is characterized by a low frequency component (of order 0.01Uinfinity/delta). In addition, the motion of the shock is found to have two aspects: a spanwise wrinkling motion and a streamwise oscillatory motion. The spanwise wrinkling is observed to be a local feature with high frequencies (of order Uinfinity /delta). Two-point correlations reveal that the spanwise wrinkling is closely related to the low momentum motions in the incoming boundary layer as they convect through the shock. The low frequency shock motion is found to be a streamwise oscillation motion. Conditional statistics show that there is no significant difference in the mean properties of the incoming boundary layer when the shock is at an upstream or downstream location. However, analyses of the unsteadiness of the separation bubble reveal that the low frequency shock motion is driven by the downstream flow.
Keywords/Search Tags:Shock, Boundary layer, Interaction, DNS, Numerical, Downstream, Motion, Low frequency
Related items