An adaptive sampling technique for optimizing the design of axial turbine endwalls

Computer-based optimization is demonstrated for the design of a 1.5 stage axial turbine casing endwall. It uses a recursive implementation of the Optimal Latin Hypercube (OLH) technique for sampling the design space and Kriging as the surrogate model for performance. The latter is evaluated by Computational Fluid Dynamics (CFD). At design and at off design conditions, reductions in the stage total pressure loss coefficient verify the effectiveness of the optimization method.