Economical experimentation via 'lean design'

Abstract

In industrial applications of design of experiments, practical constraints in resources such as budget, time, materials and manpower could lead to difficulties in completing an experiment even of moderate size. This problem can be addressed by an approach referred to as 'lean design', with which no more than a bare minimum number of response values are required, and the results of a lean design experiment can be improved incrementally up to a point identical to what can be obtained from a regular design. A numerical example illustrates both the principles and applications of this cost-effective approach to experimentation.