Separating Design Optimization Problems Into Decision-Based Design Processes

Abstract

This paper introduces the technique of separation, which replaces a design optimization problem with a set of subproblems. This separation is similar to decomposition but does not require a second-level coordination. We identify conditions under which this separation yields an exact solution and other conditions under which the error can be bounded. We show that the decision-based design framework, which seeks to find the most profitable design, can be separated into a sequence of subproblems. We also apply separation to a motor design problem and demonstrate how the surrogate constraints and objective functions affect the solution quality. These results indicate a way to apply the principles of decision-based design to design processes.