Genetically Enhanced Parametric Design for Performance Optimization Peter VON BUELOW Associate Professor, Dr. -Ing University of Michigan Ann Arbor, USA pvbuelow@umich.edu Peter von Buelow received a BArch and MS in engineering from the University of Tennessee, and a Dr.-Ing. from the University of Stuttgart. He worked in firms in the US and Germany, and now teaches at the University of Michigan. Summary ParaGen is a tool, or more accurately a method, developed over the past several years at the University of Michigan, Hydra Lab. The method uses a genetic algorithm (GA) to search for well performing geometric solutions to architectural engineering problems. The geometric solutions, which are generated with associative parametric design software, are analyzed to determine performance characteristics. The performance is then used to guide a genetic algorithm which searches for good solutions. Solutions are maintained in an online database which can be browsed, filtered and sorted by the designer as a means to both graphically and quantitatively assess the design space. The tool also accepts interactive participation by the designer in the form of ‘parent’ selection to breed new solutions based on the designer’s qualitative preference. Keywords: genetic algorithms; performance-based design; parametric geometry; exploration; optimization; parallel processing; database. 1. Introduction With the advent of associative parametric design software, such as Generative Components (Bentley Systems), Grasshopper (Robert McNeel & Associates), Digital Project (Dassault Systemes) and so on, there has been both an increased interest along with the increased capability to develop more complex, free-form geometries for architectural and engineering applications. Although these new means have exposed for the designer a wide range of form vocabulary, the rich array of geometric possibilities is usually uncoupled from performative quantities of the solutions, thus rendering the decision making process more difficult. ParaGen is intended to aid the designer by both supplying performance data along with the geometric variations, as well as providing a means of searching the design space either through programmed objective functions or by “on-the-fly” user selection. The ParaGen process works by cycling through the following steps:  Select (either with human interaction or based on a formula)  Breed (using a genetic algorithm (GA) program)  Develop (the geometry using parametric software)  Evaluate (with simulation software, e.g. FEA, Ecotect, etc.)  Rank (sort the results for exploration and selection through a web interface) Figure 1 shows a flow chart of one complete cycle. ParaGen will continue to run in this cycle until interrupted by the user or until a set number of new children have been generated without improvement (indicating stagnation).