Abstract
This paper introduces a Response Surface Model (RSM) which is constructed using both the function value and its derivative (or gradient) at the sample points. The model is a version of Interpolating Moving Least Squares (IMLS) which is essentially quadratic interpolation in a multi-dimensional table of scattered sample points. The method is especially cost-effective for response surfaces in many dimensions (on the order of 10 variables). The method is tested on a speaker assembly design problem with seven design variables, and compared to a Multiquadric RSM.
Get full access to this article
View all access options for this article.