The background of this thesis is the optimization of a diesel engine simulation at Volvo Powertrain. Proposed is an algorithm for global optimization of noisy and expensive black box function using response surfaces based on radial basis functions (RBFs). A method for RBF-based approximation is introduced in order to handle noise. New points are selected to minimize the total model uncertainty weighted against the function value. The algorithm is extended to multiple objective functions by instead weighting against the distance to the Pareto front. Numerical results on analytical test functions show promise in comparison to other (commercial) algorithms, as well as results from simulations. The algorithm also handles noisy multiple objectives, something which is lacking in previous research.
Photo credits: Nic McPhee