ABSTRACT

Solution spaces are sets of designs that meet all quantitative requirements of a given design problem, aiding requirement management. In previous works, ways of calculating subsets of the complete solution space as hyper-boxes, corresponding to a collection of permissible intervals for design variables, were developed. These intervals can be used to formulate independent component requirements with built-in tolerance. However, these works did not take physical feasibility into account, which has two disadvantages: first, solution spaces may be useless, when the included designs cannot be realized. Second, bad designs that are not physically feasible unnecessarily restrict the design space that can be used for requirement formulation.

In this paper, we present the new concept of a requirement space that is defined as the largest set of designs that (1) allows for decomposition (e.g., into intervals when it is box-shaped), (2) maximizes the useful design space (good and physically feasible), and (3) excludes the non-acceptable design space (bad and physically feasible). A small example from robot design illustrates that requirement spaces can be significantly larger than solution spaces and thus improve requirement decomposition.