A new technique for surface optimization

Abstract

A surface optimization technique has been developed for a set of experimental X-Y-Z data. Three-dimensional experimental observations are fit into a preset surface function constructed in terms of independent variables and user defined parameters. These parameters are optimized one by one until the function becomes self-consistent. A least squares criterion is used for optimization. The technique may be very conveniently used as a general program for performing multivariable linear/non-linear regression analysis avoiding matrix algebra. The technique has been successfully employed in our laboratory for testing homogeneity in standard reference materials. An application software FITSURF has been developed based on this technique.