Shape-optimization techniques are nowadays an essential tool in the design chain of turbomachinery to strike the high-performance targets demanded by modern applications. Although the geometric parametrization may affect significantly the optimization cost and ultimately the optimization outcome, the selection of control points and of the subsequent design space is usually based on heuristic considerations.

This paper proposes a cost-efficient parametrization procedure based on the ANOVA analysis of B-Spline control points. To tackle the extremely large computational burden arisen from the use of several control points, a surrogate strategy is implemented, testing four different methods. The optimization relies on surrogate-assisted evolutionary strategies coupled with an experimentally validated CFD solver.

The technique is applied to a supersonic Organic Rankine Cycle turbine cascade, which features a converging-diverging bladed channel. It is shown that only the diverging section of the suction side as well as the adjacent region of unguided turning have the major impact on the aerodynamic performance. These findings enable to select an optimal distribution of mobile control points in the optimization block, leading to significant savings in computational cost.

Finally, three optimizations are carried out, varying locally the number of control points; results are widely discussed in terms of both optimization outcomes and optimizer robustness.

This content is only available via PDF.
You do not currently have access to this content.