Adaptive radial basis function emulators for robust design

In the field of engineering design, tradeoffs between competing design objectives can only be made if there is a good understanding of the product or process under development. To facilitate this, adaptive classes of models can be used to represent complex engineering systems and provide important information for design development. This paper describes a fast implementation of a radial basis function model, intended for this purpose. By exploiting the mathemtical form of the Gaussian basis function, a computationally efficient method of estimating smoothness is developed and used in the model fitting process. The method is applied to a set of existing experimental data and compared with two alternative modelling strategies involving polynomial and stochastic process models.