The use of dimensional analysis to augment design of experiments for optimization and robustification

Abstract

Optimization can be a time-consuming and demanding process when an analytical model of the system of interest cannot be developed. The problem is even more extreme when the robustification of the system is desired. Under such circumstances the design engineer will typically resort to design of experiments (DOE) or finite element analysis/computational fluid dynamics or some combination. Each can be time consuming and demanding. It is shown in this paper that by combining dimensional analysis with DOE it is possible to generate a near-exact surrogate model of a system empirically. This can be done with a significantly reduced number of experiments when compared with traditional DOE techniques. The approximation is sufficiently accurate to be optimized or robustified using methods traditionally suited to analytical models. A strategy designed to help the design engineer take full advantage of this approach is presented.