Assessing the Risk of Space Reduction Decisions in Set-Based Design through Fragility-Tracking of Interdependent Design Spaces

Abstract

As designers execute analyses and explore potential design solutions, their understanding of a design problem grows. They begin to learn the trade-offs of pursuing certain design solutions over others, and they form preferences that are grounded in how they perceive design solutions to satisfy each discipline's requirements. Once designers start making decisions from these perceptions, they begin to limit the knowledge generation associated with their decisions. If they are not careful, their decisions will prevent new knowledge from integrating with existing knowledge, and they will incur emergent design failures. Emergent design failures incited by iterative decisions are less consequential to designers because rework cycles are naturally apart of their learning process. Emergent design failures incited by convergent decisions, on the other hand, are more consequential to designers because they expend substantial time and effort exploring design spaces before making any decisions in the first place.

Without the ability to anticipate whether elimination decisions will incite emergent design failures, convergent design approaches become far less appealing. The benefits of adopting a convergent approach over an iterative one are well-documented, especially as design problems become increasingly complex, but why undertake the added effort if these decisions can be so much more detrimental to attaining a feasible outcome? To encourage designers to carry out a convergent design approach, they need a way to assess how vulnerable the perceptions of feasibility in their design spaces are to being invalidated by new information before committing to a space reduction decision. The Probabilistic Fragility Model (PFM) and Entropic Fragility Model (EFM) introduced in this work are intended to do just that. Both fragility models assess the fragility of a reduced design space relative to its non-reduced design space and quantify the risk of incurring emergent design failures when going through with a space reduction compared to delaying the space reduction.

The utility of the frameworks are evaluated by executing a series of convergent, set-based design (SBD) simulations for a polynomial design problem and a bulk carrier design problem with and without them and observing the emergent design spaces. The PFM's process is a bit more straightforward and is able to reasonably delay space reductions for the polynomial design problem more than the bulk carrier design problem without significantly driving up the added cost of retaining more solutions. The EFM is able to reasonably delay space reductions for both problems without significantly driving up the added cost, but at the expense of being a more involved process than the PFM. After testing both frameworks out on the polynomial design problem, their initial versions still lacked the capability to identify design space fragilities from some more niche sources of new information. So before testing out the frameworks on the bulk carrier design problem, three extensions are proposed and incorporated to expand their reliability. The findings of this work show that making fragility checks with the frameworks is a very worthwhile precaution for designers to take to help them avoid space reduction decisions that incite emergent design failures.