I am validating an airfoil at Re=6 million at mach 0.32 within FLUENT. I have found over the five grids examined thus far that a grid with approx. 92,000 cells at y+ about 150-300 produces lift and drag results that are within 3-4% of experimental data, over an AOA range of -1 to 12 deg.

Since i am using the SST k-w model, i would like to further test result convergene by increasing the number of cells such that y+ of 1 can be obtained. I am limited in my computer resources so have been able to achieve a y+ of 25, yet my results are now worst off, than they were with a y+ of 150-300. I have maintained a cell grid aspect ratio of ~20 within the BL, and have about 340 grid points on the airfoil surface. I am confused as to why my results are worst off now; i would have expected similar results or slight improvement given that the y+ has been reduced to capture the visoucs sub-layer. Would they deteoriate even further if i was to achive a y+ of 1? Any advice???

The fact that you're getting reasonable drag results with a very high y+ of around 200 implies that the turbulence model is using wall functions. Depending on the approach, the model may actually require a reltively high y+ in order to work (e.g. some wall functions methods require that the first grid point off the wall is within the log region of the boundary layer). Furthermore, depending on the method, not only the first grid cell is important, but also how many cells fall within the boundary layer. You better read the manual and find out what method Fluent is using, what options you have, and what the requirements are on grid resolution and y+.