Multi-fidelity design optimization of Francis turbine runner blades

Abstract

A robust multi-fidelity design algorithm has been developed, focusing to efficiently handle industrial hydraulic runner design considerations. The computational task is split between low- and high-fidelity phases in order to properly balance the CFD cost and required accuracy in different design stages. In the low-fidelity phase, a derivative-free optimization method employs an inviscid flow solver to obtain the major desired characteristics of a good design in a relatively fast iterative process. A limited number of candidates are selected among feasible optimization solutions by a newly developed filtering process. The main function of the filtering process is to select some promising candidates to be sent into the high-fidelity phase, which have significantly different geometries, and also are dominant in their own territories. The high-fidelity phase aims to accurately evaluate those promising candidates in order to select the one which is closest to design targets. A low-head runner case study has shown the ability of this methodology to identify an optimized blade through a relatively low computational effort, which is significantly different from the base geometry.