Exploring Methods to Quantitatively Compare Optimization Techniques for Building Envelope Design

Building designers often face tradeoffs when comparing designs that are not readily reducible to single objective functions, but instead benefit from evaluating multiple objectives, this is especially true for building envelope design. Multiple Objective Optimization (MOO) methods are available to assist designers to systematically search through large numbers of design
alternatives to identify high-performing design solutions in terms of two or more objectives. This paper evaluates three genetic algorithms (GAs) and a gradient-based algorithm that designs various elements of a building's envelope attempting to minimize both life-cycle costs (LCCs) and environmental impact. We evaluated the solution quality of each algorithm in
terms of multiple performance indicators, and various convergence criteria. The results show that the Darwin and DAKOTA GAs are consistently top performers. The evaluation also demonstrates the value of the hyperarea as a dual objective indicator of solution quality and synthesizes much of the available literature on the metrics of multiobjective solution quality.