Multi-fidelity Gaussian Process Regression for Propeller Optimisation Under Uncertainty

In this work, an optimisation workflow is presented for uncertainty-based design optimisation using a multi-fidelity Gaussian Process Regression. The motivation of the proposed techniques is to solve a propeller design optimisation problem under uncertainty. This problem involves expensive CFD simulations to evaluate the aerodynamic performance which prohibits the application of standard optimisation techniques and the direct calculation of statistical measures. Therefore, a surrogate-based optimisation workflow is presented. The computational budget limits the number of high-fidelity simulations which makes impossible to accurately approximate the design landscape. This motivates the use of cheap low-fidelity simulations to obtain more information about the unexplored locations of the design landscape. The information stemming from the low- and high-fidelity numerical experiments can be fused together with multi-fidelity Gaussian Process Regression to build an accurate surrogate model despite the low number of high-fidelity simulations. Although Gaussian Process can inherently model uncertain processes, here the design and uncertain parameters are treated separately and only the design space is modelled by a Gaussian Process. The probabilistic space is modelled by a Polynomial Chaos Expansion. The combination of the above techniques allows us to efficiently carry out a design optimisation problem which otherwise would be impractical.