Comparison of Several Optimization Strategies for Robust Turbine Blade Design

Abstract

This paper addresses the problem of turbine blade shape optimization in the presence of geometric uncertainties. Several strategies are tested and compared on a two-dimensional compressor blade optimization process for which performance is assessed using a commercial Reynolds-averaged Navier―Stokes computational fluid dynamics code. In each case, a range of shape errors are considered that attempt to simulate foreign object damage, erosion damage, and manufacturing errors. These lead to stochastic performance measures that, in turn, are considered in a multi-objective optimization framework. Because of the long run times associated with Reynolds-averaged Navier―Stokes codes, use is also made of surrogate or response-surface-based optimization methods to speed up the search processes. The paper shows that a range of techniques can be used to tackle this problem, but that no one method is clearly best overall. The practitioner is therefore cautioned against favoring a single approach for such design problems. Further research may help clarify these issues.