authors:

Mario Moacir Deuss, Mark Pauly (Dir.)

links:

EPFL PhD Thesis No. 6685

Architectural structures such as buildings, towers, bridges and roofs are of fundamental importance in urban environments. Their design, planing, construction and maintenance carry numerous challenges in engineering due to the complex interplay of material, form, spaces and statics. The history of architecture shows proof of the many approaches humanity has come up with to tackle those challenges. To mention a few, think of tents, timber huts, pyramids, masonry bridges, steel structures and modern skyscrapers. The methods presented in this thesis are tailored to handle the general case of freeform architectural structures. A common way to develop novel architectural structures is to produce prototypes at various scales. Physical prototypes, however, do not allow for quick changes of aspects of a design such as material and form. The advent of computer-aided design tools alleviated some of these limitations, but brought with it new challenges in terms of simulation of physics and interaction with virtual content. Compared to physical prototypes, digital ones can in theory include many more types of constraints by leveraging numerical computation. To be practical, however, a digital prototyping tool needs to be designed carefully considering efficiency, generality, accuracy, simplicity and robustness of its implementation. Also, there are currently many unsolved problems in the digital exploration of desirable and feasible designs with respect to constraints. Important constraints are imposed directly or indirectly by the ease and cost of realization and maintenance of a freeform architectural structure. For example, the geometry of components making up a structure can have a big impact on the cost of fabrication: planar components can simply be cut out of material that usually comes in flat sheets, while curved ones tend to require more costly production processes. The construction and assembly of architectural structures can cause a considerable part of the full cost. A well-chosen assembly sequence can reduce both labor and necessary temporary support structures such as scaffolds. This thesis approaches challenges in the context of computational prototyping tools for architectural structures by studying three concrete subproblems and proposing practical solutions to them. We present a constraint-aware modeling tool capable of robustly simulating physics and handling various geometric constraints at interactive rates. We show how constraints relevant to aesthetics and production can be implemented in our tool, yielding a fabrication-aware modeling tool. We present a computational method for finding a mass-producable approximation of a given surface which minimizes fabrication cost. The method optimizes for a set of molds each of which can be used to produce multiple components, so-called panels, of the approximation, while respecting user-defined constraints on the continuity and deviation. The problem this method solves is referred to as paneling, which in turn is an instance of the rationalization problem: approximating input under constraints relevant to the physical realization of a structure. We present a computational method for minimizing the work necessary for the construction of a freeform self-supporting structure. We study the use of chains to support the structure during assembly. Our method searches for an assembly sequence of the structure's components which minimizes the number of times a chain has to be rehung.