Turbine blades can be optimised to reduce aerodynamic losses which therefore improves engine efficiency and in particular adiabatic efficiency. This has readily been done by deterministic optimisation, however Kamenik presents an alternative robust optimisation procedure which utilises probability distributions (usually Gaussian). This has the advantage of dealing with uncertainties in manufacturing and therefore prevents initial constraints being violated.
Robust optimisation works by considering the mean and variance together of the probability distributions from many data samples against an objective function, say adiabatic efficiency, and then a Pareto Front is used to decide on the optimal solution.
Kamenik has implemented noise factors into his CFD and the results from these have been utilised in Monte Carlo methods on a meta model. Monte Carlo methods in this case involve randomly sampling the design space to obtain numerical solutions. This optimisation is then solved once the Pareto Front converges.