A novel methodology is developed to integrate state-of-the-art computationa
l fluid dynamics analysis, NURBS, and optimization theory to reduce total p
ressure distortion and sustain total pressure recovery within a curved thre
e-dimensional subsonic S-duct diffuser by automated redesign of the diffuse
r shape. Two independent design variables are used. The change of the surfa
ce shape is assumed to be Gaussian. GASP with the modified Baldwin-Lomax tu
rbulence model (Baldwin, B. S., and Lomax, ii., "Thin Layer Approximation a
nd Algebraic Model for Separated Turbulent Flows," AIAA Paper 78-257, 1978)
is employed for the flowfield prediction and proved to give good agreement
with the experimental surface pressure for the baseline S-duct diffuser ge
ometry. The automated design optimization is performed with a gradient-base
d method to minimize the total pressure distortion based on the two design
variables. The best configuration obtained reduced distortion by typically
70% while keeping the total recovery essentially the same. The results indi
cate that the mechanism responsible for improved diffuser performance is th
e suppression of detrimental secondary Rows by changing the surface shape t
o redirect the flow.