Design Space Exploration for Quantifying a System Model’s Feasible Domain

Abstract

A major challenge in multidisciplinary system design is predicting the effects of design decisions at the point these decisions are being made. Because decisions at the beginning of system design, when the least is known about the new system, have the greatest impact on its final behavior, designers are increasingly interested in using compositional system models (system models created from independent models of system components) to validate design decisions early in and throughout system design. Compositional system models, however, have several failure modes that often result in infeasible or failed model evaluation. In addition, these models change frequently as designs are refined, changing the model domain (set of valid inputs and states). To compute valid results, the system model inputs and states must remain within this domain throughout simulation. This paper develops an algorithm to efficiently quantify the system model domain. To do this, we (1) present a formulation for system model feasibility and identify types of system model failures, (2) develop a design space exploration algorithm that quantifies the system model domain, and (3) illustrate this algorithm using a solar-powered unmanned aerial vehicle model. This algorithm enables systematic improvements of compositional system model feasibility.