Philip E. Morgan

Large-Eddy Simulation of Separation Control for Flow over a Wall-Mounted Hump

This work describes an implicit large-eddy simulation (LES) for active control of flow over a wall-mounted hump. Both steady suction and oscillatory blowing and suction are compared to baseline conditions with no flow control. This simulation models an experiment conducted by NASA which was one of the test cases in their 2004 CFD Validation on Synthetic Jets and Turbulent Separation Control Workshop. A previous LES of the baseline case with the current scheme demonstrated significantly better agreement with the experimental flow physics than RANS in the separated region downstream of the hump. The current work concludes that this also holds for the cases using active control. The LES is accomplished using an implicit parallel flow solver that is based on an approximately-factored time-integration method using fourth-order spatial compact-dierence formulations and a high-order filtering strategy. To properly resolve the flow for this LES, the Reynolds number of 9.36 ◊ 10 5 used in the experiment was reduced to 2.0 ◊ 10 5 . Eects of lowering the Reynolds number were previously investigated using RANS. Flow features of the active control cases are compared with the baseline case, each other, and experimental data. The active control LES cases matches experimental data significantly better in the recirculation region than RANS. The baseline and steady suction LES separation reattachment lengths are within 2% and 4%, respectively, of the experimental locations. These simulations achieve very good agreement with the experimental surface pressure coecient, skin friction coefficient, mean velocity profiles, Reynolds stresses, and flow reattachment location. Because of the lower Reynolds number, the oscillatory blowing and suction flow control only has about 25% of the eectiveness of that observed in the experiment. The LES separation reattachment length is 10% longer than its experimental counterpart. Other flow quantities display favorable agreement with experimental data. Results demonstrate that both steady suction and oscillatory blowing and suction can be properly simulated by LES and eectively reduce the size of the separated flow region.

Numerical Simulations Investigating Control of Flow Over a Turret

Hybrid Reynolds Averaged Navier-Stokes/implicit Large Eddy (hybrid RANS/ILES) simulations were performed for flow over a turret to compare the effectiveness of three forms of flow control: oscillatory blowing/suction, steady suction through a slot, and lowvelocity suction through a porous surface. The hybrid RANS/ILES computations were obtained using a well-validated high-order solver employing a 4-order compact spatial discretization in conjunction with a 6-order low-pass spatial filter operator that regularized the evolving solution by selectively removing unresolved high-frequency content. The turret configuration consisted of a half-foot radius hemisphere mounted on a 4.5′′ tall circular cylinder attached to the wind tunnel wall. The flow conditions were M∞ = 0.4 and ReD = 2.4× 10. Numerical simulations were performed on both a 14 million and 23 million point mesh. Results from the numerical simulations are compared with previous numerical and experimental data for a baseline case without flow control. The simulations demonstrate that steady suction through both the slot and the porous surface is more effective in delaying flow separation from the turret dome than oscillatory blowing and suction for the frequency and amplitude chosen. Delaying separation significantly reduced turbulent kinetic energy and small scale flow structures in the turret dome wake which degrade and distort optical wavefronts. Both types of steady suction flow control generated attached flow over the entire dome of the turret. Steady suction through the porous turret also significantly reduced the size of the wake vortices. Additional research is needed to improve the effectiveness of the oscillatory flow control.

High-order numerical simulation of turbulent flow over a wall-mounted hump

The development of a high-order spatial discretization for a k-e turbulence model and its application to flow over a wall-mounted hump is described. The high-order implementation is validated for a flat plate and subsequently applied to the more complex wall-mounted hump for conditions with and without flow control. Results for the hump flow are compared to experimental data. The turbulence model is incorporated in an implicit parallel flow solver that is based on an approximately factored time-integration method coupled with spatially fourth- and sixth-order compact-difference formulations and a high-order filtering strategy. Both second-order and high-order discretizations of the k-e turbulence equations were included in the compact solver. Validation using flow over a flat plate demonstrated that use of a second-order scheme for the k-∈ turbulence equations dominates the solution even when high-order compact differencing is used for the flow equations. This validation also demonstrated that significant computational savings are possible because less mesh resolution is required when using a high-order discretization of the k-e turbulence equations. Comparison of the high-order and second-order solutions was also performed for the wall-mounted hump. Qualitative agreement was achieved with experimental data for both high-and low-order schemes. High-order solutions on a coarse grid agreed very well with second-order solutions on a considerably finer grid.

Hybrid Reynolds-Averaged Navier-Stokes/Large-Eddy Simulation Investigating Control of Flow over a Turret

This investigation explores the flow over a turret atM1 0:4 and ReD 2:4 10 using both hybrid Reynoldsaveraged Navier–Stokes/implicit large-eddy simulation (RANS/ILES) and k-based unsteady Reynolds-averaged Navier–Stokes (URANS) simulations. Additional hybrid RANS/ILES simulations were performed to compare the effectiveness of two steady-suction flow control approaches. The hybrid RANS/ILES computations were obtained using awell-validated high-orderNavier–Stokes flow solver employing a fourth-order compact spatial discretization in conjunction with a sixth-order low-pass spatial filter. The URANS simulations were performed using a secondorder version of the flow solver and kturbulence model. The turret configuration consisted of a half-foot radius hemisphere atop a 4.5-in.-tall circular cylinder base. Both steady suction through a slot and a leeward porous turret shell were explored as forms of flow control. Time-mean hybrid RANS/ILES results, obtained on a 23 10-point mesh, compared reasonably well to experimental pressure coefficient and velocity profiles for the baseline flow. The separation angle was predicted to within 3 deg of experimental observations. The instantaneous hybrid RANS/ILES solutions display complex three-dimensional flow phenomena in the wake of the turret that the kURANS model was unable to resolve. Both steady-suction flow control approaches successfully attached the flow over most the turret dome and significantly reduced the size of the wake.

Large Eddy Simulation of Flow Over a Flat-Window Cylindrical Turret with Passive Flow Control

This work presents multiple high-order implicit large eddy simulations (ILES) of flow over a cylindrical turret with a flat window oriented at three angles (90, 100, and 120 degrees). For the 100 degree case, an additional computation was performed with a row of five pins inserted upstream of the turret to passively control flow separation. The ILES computations were obtained using a well-validated high-order Navier-Stokes flow solver employing a 6-order compact spatial discretization in conjunction with a 8-order lowpass spatial filter. Simulations were executed on a massively parallel computing platform using a high-order overset grid methodology on meshes with approximately 55 million nodes. Results from this simulation were compared to data from an experimental study performed at the University of Notre Dame. Overall, solutions compared favorably to experimental time mean and fluctuating velocity profiles as well as the overall flow structure at all three angles. Pins inserted upstream of the flat aperture generate spanwise structures in the time-mean flow that energize the boundary layer and reduce the size of the separation bubble at the leading edge of the flat window. Reducing the size of this separation region led to an overall decrease in the turbulent kinetic energy over the remainder of the window. Additionally, flow sheds from the pins at a discrete frequency similar to that observed in another LES simulation for a finite-height cylinder. This study demonstrates the ability of ILES to successfully reproduce complex flow physics over a flat-window turret with various levels of separation. Additionally, it provided valuable insight into the flow physics of pins and how they control flow separation which is critical in reducing aero-optical aberrations.

Numerical Investigation of Flow Around a Turret

k turbulence model. The flow conditions for the turret simulation were M1 = 0.5 and ReD = 4.36 ◊ 10 5 . Results from hybrid RANS/ILES simulation are compared to the RANS solutions and experimental data. The hybrid RANS/ILES solution displays complex 3-D flow phenomena in the wake of the turret that the k RANS model is unable to resolve. The numerical simulations indicate that turbulent separation occurs from the top of the turret instead of the experimentally observed laminar separation. Additional computations employing an implicit LES are needed to determine if the k turbulence model, used in both the hybrid RANS/ILES and RANS simulations, is responsible for the turbulent flow separation from the turret dome.

Hybrid RANS/LES Simulations for Flow Around Two Turret Configurations

Both hybrid Reynolds Averaged Navier-Stokes (RANS)/ implicit Large Eddy Simulation (ILES) and k− based unsteady RANS (URANS) simulations were performed for flow over two turret configurations. The hybrid RANS/ILES computations were obtained using a well-validated high-order solver employing a 4-order compact spatial discretization in conjunction with a 6-order low-pass spatial filter operator that regularized the evolving solution by selectively removing unresolved high-frequency content. The URANS simulations were accomplished using a k− turbulence model with a 2-order Navier-Stokes flow solver. The first configuration modeled an experiment of a 1 1 2 ” turret at flow conditions of M∞ = 0.5 and ReD = 4.36× 10. The second configuration modeled a one-foot turret at flow conditions of M∞ = 0.4 and ReD = 2.4 × 10. URANS and hybrid RANS/ILES simulations were performed on both configurations using a coarse four-million point mesh. Hybrid RANS/ILES simulations were also conducted on a fine 23 million point mesh. Results from the numerical simulations are compared with experimental pressure, velocity and surface-flow visualization. Independent of the mesh resolution and approach, the numerical simulations for the first turret configuration predicted separation from the turret dome downstream of the experimental location. This result suggests that URANS and hybrid turbulence models were unable to properly simulate the laminar/transitional separation. The simulations at higher Reynolds numbers for the second turret configuration yielded much better agreement with experimental data. The fine-mesh hybrid RANS/ILES solution agreed well with experimental pressure coefficient profiles and was within 3 of the separation angle. For both configurations the hybrid RANS/ILES solutions display complex 3-D flow phenomena in the wake of the turret that the k− URANS model was unable to resolve. The extent that these resolved flow structures degrade and distortion optical wavefronts will be the subject of future research.

Large-Eddy Simulation of Flow Over a Wall-Mounted Hump

formulations and a high-order filtering strategy. Results for the hump flow are compared to Reynolds Averaged Navier-Stokes (RANS) simulations and experimental data from a recent NASA workshop on synthetic jets and turbulent separation control. Participants at this workshop using RANS, hybrid RANS/LES, and direct numerical simulations (DNS) achieved only qualitative agreement with experimental data. The current work attempts to expand on that knowledge base by correcting for wind tunnel blockage and removing the reliance on turbulence models which are known to perform poorly in flows with separation. To properly resolve the flow for this LES, the Reynolds number of 9.36 ◊ 10 5 used in the experiment was reduced to 2.0 ◊ 10 5 . Eects of lowering the Reynolds number were investigate using RANS. At this lower Reynolds number, the LES results displayed a slightly dierent character than the experiment in the skin-friction coecient over the front portion of the hump and in the velocity profiles near the separation point on the hump. In the recirculation region, the LES results matched experimental data significantly better than RANS. The LES properly modeled the flow physics of the separation bubble achieving better agreement in the surface pressure coecient, skin friction coecient, mean velocity profiles, Reynolds stresses, and flow reattachment location.

Numerical Simulation of Flow Over a Submerged Hemispherical Flat-Window Turret

Highdelity implicit large eddy simulation (HFILES) have been performed for ow over a 3-D at-window hemispherical turret mounted on a surface for both the baseline ow and for the case of steady suction through a slot at the edge of the aperture to control separation. The on-coming boundary layer is 45% the height of the turret. The diameter of the at-window aperture is 48% of the turret diameter and is oriented symmetric to the oncoming ow at an elevation angle of 138. The ow conditions were M = 0:10 and ReH = 50; 000. This geometry is modeled after an experiment performed at the University of Florida. The HFILES computations were obtained using a well-validated high-order NavierStokes ow solver employing a 6-order compact spatial discretization in conjunction with a 8-order low-pass spatial lter. The simulations were performed on both 57M and 127M point mesh systems using a massively parallel computing platform and a high-order overset grid methodology. Qualitative comparisons are made with the experiment and previous simulations of a at-window hemispherical turret on a cylindrical base. Overall, solutions compared favorably to experimental time mean solutions and oilow visualization. This study demonstrates that steady suction at the edge of the aperture can be successfully employed to signi cantly reducing ow separation at large look angles in the wake of the turret which is critical in reducing aero-optical aberrations.

Numerical Simulation Investigating Control of Flow Over a Flat-Window Turret

This investigation explores the ow over a at-window turret at M1 = 0:35 and ReD = 1:98 10 using a hybrid Reynolds Averaged Navier-Stokes/highdelity implicit Large Eddy Simulation (hybrid RANS/LES). Hybrid RANS/LES simulations are performed for both the baseline turret and the turret with a steady-suction slot around the edge of the atwindow aperture to control ow separation. The hybrid RANS/LES computations were obtained using a well-validated high-order Navier-Stokes ow solver employing a 6-order compact spatial discretization in conjunction with a 8-order low-pass spatial lter. The turret con guration consisted of a half-foot radius hemisphere atop a 400 tall circular cylinder base. The orientation of the 5:400 at window is symmetrical to the upstream ow at an elevation angle of 110. Results show that the at window introduces massive ow separation over the aperture not seen on a previously studied conformal window turret. Other time-mean and instantaneous ow characteristics are described. Steady suction through the slot is shown to be e ective in virtually eliminating separation over the at-window aperture thus removing signi cant ow structures that lead to aero-optical aberrations.