The surrogate assisted evolution of aerofoil geometries
(a type of blade) has been widely explored for use with aircraft
design. Some examples include, Giotis and Giannakoglou [66]
who used multiple output neural networks as surrogate models
for multiobjective aerofoil optimization. Emmerich and
Naujoks [67] and Kumano et al. [68] used kriging to provide
approximations for multiobjective aerofoil design. In addition,
Zhou et al. [69] evolved aerodynamic aerofoil geometries with
a representation consisting of Hicks–Henne bump function
parameters. The EA was assisted by both a global and local
surrogate model. Significantly, these approaches use simulations
to evaluate candidate solutions and typically consider
only 2-D blades (due to the cost of CFD analysis).