NGC

The NGC project is a continuation of at least one other project sponsored by Northrup Grumman Corporation studying active flow control in ducts with various geometries that tend to induce large adverse pressure gradients and massive separation in the flow. The present project concerns high subsonic flow in a square channel which is subjected to a aggressive diffuser. Flow control is performed with unsteady tangential blowing near the upstream side of the diffuser. The previous project was an aggressive S-duct which used similar tangential blowing to keep the flow attached.

The CFD at CU has been conducted in conjunction with experiments at RPI for several years. In that time, the primary student working on the CFD has changed from Yi Chen to Kyle Woolwine to Nicholas Mati. It is the intent of this page to provide some continuity for future transitions with brief documentation of present work. It is also hoped that maintaining this page will increase communication with Prof. Jansen, and help keep track of simulation cases as they proliferate.

---

S-Duct

This page was created after Yi stopped working on the project. For more information, read Yi Chen's thesis.

Diffuser

Series 10

Series 10.3

The only difference between series 10.2 and 10.3 is the splitting of a surface in the blower near the lip into three separate surfaces. This allowed the mesh height on the lower blower wall to be manually cut down near the lip which fixed an issue from 10.2 where the mesher was not trimming the boundary layer correctly in a refinement box for some reason.

-Interpolating from the 10.1 solution to the M3 mesh, high z-velocity oscillations clipping at +-250 m/s continued in the blower. This solution was abandoned.

-Interpolating from the 11.2 solution to the M3 mesh, the blower started off with nearly zero velocity. Some inflow was initially observed in the blower, but no substantial z-oscillations were observed. However, backflow was present in the 11.2 solution which resulted in temperature clipping high at the bottom of the diffuser with at most 2 orders of magnitude of convergence in the nonlinear residual. Dropping the upper limit from 400 K to 340 K to 320 K resulted in only minor benefit and the solution was abandoned.

-Starting from near zero velocity initial conditions on the M3 mesh broke phasta. There was a version compatibility issue with the smd which resulted in all flow fields being zero. While debugging, the geometry was remeshed.

-Starting from near zero velocity initial conditions on the M4 mesh resulted in similar flow separation and temperature clipping. Elevated viscosity of 10 nu, 100 nu, and 1000 nu was tried in the diffuser, but this resulted in only minor improvement.

-M5 actually recycled the mesh from M4. The main difference was with boundary conditions on the diffuser slip walls. The turbulence wall designation was removed and scalar 1 = 0 was changed to scalar 1 flux = 0. Slip boundary conditions were finally observed, although back flow was still present at the outlet which caused the solution to diverge.

-M6 again recycled the mesh from M4, but swapped the homogeneous natural temperature and scalar BCs at the outlet for essential BCs. This created a large gradient on elements touching the outlet, but the velocity was generally high enough to prevent substantial changes further upstream.

It was also observed that the wall temperature was still being set to 317 K, despite requesting a zero heat flux BC in the GUI. A quick grep through the code yielded a hack in gendat.f that Yi and introduced which set the temperature, all three components of velocity, and scalar 1 whenever a wall was designated as a turbulence wall. I [Nicholas] commented most of the offending code, but kept the scalar 1 = 0 BC. At least for the SA turbulence model, scalar 1 = 0 and turbulence wall should always accompany one another, although this might still cause issues if someone tries to use the k-w model.

Roughly 2000 time steps of RANS were run with the new boundary conditions. Small amounts of backflow were present, but the solution generally appeared to be fine in the diffuser. High z-velocity oscillations were again observed in the blower and were likely responsible for the divergence of the solution. Another 1000 steps were run with DDES, but again, the solution appeared to diverge near the blower, even when the blower was turned off.

-Under the hypothesis that the large gradient in mesh size between the bottom of the blower and the fillet was responsible for introducing the high z-velocities, the blower was refined by a factor of between 2 and 4 to produce the M7 case.

Series 12

The Series 12 case is a hybrid between the series 8 and 10 cases; the same blower geometry is derived from series 10 but the diffuser is removed as in series 8.