Generative Components and Genetic Algorithms
Genetic algorithms aim to mimic natural selection in the design process. A set of parameters or “genes” characterize a “species” of artifact. Individuals within the species express different values for those genes. A fitness function evaluates each individual’s health. The algorithm works by assigning random gene values for several individuals, evaluating them, discarding the weakest ones, breeding the strongest ones by interchanging genes, and repeating for successive generations. Genetic algorithms sometimes yield surprising designs that a strictly deductive deterministic design process might not discover.
This project uses Bentley Generative Components to script parametric designs for several classes of structures, including folded plates, branching columns, and geodesic domes. Bentley STAAD structural analysis serves as the fitness function.
Monica Ponce de Leon (Dean of Architecture and Urban and Regional Planning) is the principal investigator. Peter von Bülow (Associate Professor of Architecture) develops the genetic algorithms. Ted Hall worked with recent Architecture graduates Jason Dembski and Kevin Deng to script the structures and visualize them at full scale 3D in the MIDEN.