Learning to be a Vault - Implementing learning strategies for design exploration in inter-scalar systems
David Stasiuk, Mette Ramsgaard Thomsen · eCAADe proceedings · 2014
Parametric design models enable the production of dynamic form, responsive material assemblies, and numerically and geometrically analytical feedback. The value potential for design produced through the procedural transformation of input parameters (or features) through algorithmic models has been repeatedly demonstrated and epistemically refined. However, despite their capacity to improve productivity and iteration, parametric models are nonetheless prone to inflexibility and reduction, both of which obscure processes of invention and discovery that are central to an effective design practice. This paper presents an experimental approach for the application of multiple, parallel computational design modelling strategies which are tested in the production of an inter-scalar model array that synthesises design intent, the simulation of material behaviours, performance-driven adaptation, and open-ended processes of discovery and categorical description. It is particularly focused on the computational potentials embedded in interdependent applications of simulation and machine learning algorithms as generative and descriptive drivers of form, performance, and architectural quality. It ultimately speculates towards an architectural design modelling method that privileges open model topologies and emergent feature production as critical operators in the generation of flexible and adaptive design solutions.