Abstract
We introduce the concept of robust design in the context of high-low tolerancing and propose a mathematical programming model for it. In situations where the transfer function is available either in closed form or through computer simulation, this model can be solved using conventional non-linear programming techniques. We also introduce an approximation procedure for solving this problem experimentally. This procedure can be used in situations where either the transfer function is not available in closed form or the variables are discrete or categorical. A numerical example and two case studies from the open literature are solved using these techniques and results are presented.