Synthesizing design performance: An evolutionary approach to multidisciplinary design search

Abstract

Design is a goal-oriented decision-making activity. Design is ill defined and requires synthetic approaches to weighing and understanding tradeoffs amongst soft and hard objectives, and imprecise and/or computationally explicit criteria and goals. In this regard, designers in contemporary practice face a crisis of sorts. How do we achieve performance under large degrees of uncertainty and limited design cycle time? How do we better design for integrating performance? Fundamentally, design teams are typically given neither enough time nor the best tools to design explore, generate design alternatives, and then evolve solution quality to search for best fit through expansive design solution spaces. Given the complex criteria for defining performance in architecture, our research approach experiments upon an evolutionary and integrative computational strategy to expand the solution space of a design problem as well as presort and qualify candidate designs. We present technology and methodology that supports rapid development of design problem solution spaces in which the objectives of three design domains have multidirectional impact on each other. The research describes the use of an evolutionary approach in which a genetic algorithm is used as a means to automate the design alternative population as well as to facilitate multidisciplinary design domain optimization. The paper provides a technical description of the prototype design, one that integrates associative parametric modeling with an energy use intensity evaluation and with a financial pro forma. The initial results of the research are presented and analyzed including impacts on design process; impacts on design uncertainty and design cycle latency; and the affordances for “designing in” performance and managing project complexity. A summary discussion is developed that describes a future cloud implementation and the future extensions into other domains, scales, tectonic, and system detail.