Improving genetic algorithm for design optimization using architectural domain knowledge

Abstract

The integration of building performance simulation and design optimization in the early stages of the architectural design process has attracted a high volume of research in recent years. However, both building simulation and design optimization require a significant amount of computing time, especially when there are multiple design objectives to achieve. In this paper, we present a technique–offline simulation–to effectively reduce the computing time in such design optimization problems. The validation of this method is presented in the context of a case study with parametric form-finding for a nursing unit design with two design objectives: minimizing the nurses’ travel distance and maximizing daylighting performance in patient rooms. The results show that computing time can be reduced significantly during the simulation and optimization process. The technique presented is based on Genetic Algorithm (GA). The use of GA in architectural design has become a trend for design optimization. Currently, however, only the general method of GA is applied to architectural problems. This research provides a new type of study that utilizes architectural domain knowledge to customize GA techniques in order to significantly improve the design optimization process.