Building the Digital World – Architectural Design Methods Based on the Use of Digital Tools – Performance Based Form Generation and Optimization

Grobman Yasha
Ph.D.Thesis, 2008

(together with Yezioro Abraham)


There has already been one loss of innocence in the recent history of design; the discovery of machine tools to replace hand craftsmen … now we are at a second watershed. This time the loss of innocence is intellectual rather than mechanical. (Christopher Alexander, 1964)

The introduction of computers to architectural design has been making a significant impact on the way buildings are being designed and built. For the second time in the history of modern design, technology is advancing faster than the building industry. This time, the digital and information technology (IT) revolutions introduced technologies that allowed for the development of architectural designs and manufacturing tools on grounds other than need, and thus their influence on architecture

is still largely unknown. This research examines changes in the architectural design process caused by the introduction of computers focusing mainly on computer-based form generation, simulation and evaluation. It suggests a new generative performance oriented design (GenPOD) model that use “performance envelopes” in a non linear generative design method.

As a preliminary stage a digital architecture database was developed. The database, which included projects that followed a new definition for digital architecture or computer-oriented design, has shown trends in computer oriented design project types, costs, geographical distribution and other details starting from the mid 1990’s.

Following the preliminary stage the research focuses on developments in architectural design tools/software, design methods and computer-based generation, simulation and evaluation tools and approaches. It discusses the increase in level of control that architects have over the designed architectural form, which is based on the increase of the amount of data the architectural form embeds, and its implications on the architectural design process. It also discusses, via a division into form- and performance based design, the shift towards performance-based architectural design, which introduced new possibilities in terms of using computers in the design process. The new possibilities derive, among other, from the ability to embed/add empirical quantitative data to the geometrical information regarding the architectural form and negotiate quantitative and qualitative data in the generation and evaluation processes of the initial architectural design.

Based on these analyses the research introduced the notion of multiple performance envelopes as a base for the GenPOD model. Performance envelopes are surfaces that connect points with similar information regarding performance (E.g. wind performance envelope will be defined by all the points in the design space with a similar wind speed). As opposed to traditional approaches in which computer simulation is used in an “after the fact” manner in order to evaluate architectural form fulfillment of certain performative demands, the new model suggests to use performance envelopes in a generative, “before the fact” approach.

The suggested model allows generating architectural initial form using a negotiation process of one or more performance envelopes that stand for similar or different performances, regarding different aspects of the design. The generation process is parametric and iterative in a sense that it allows to combine numerous rounds of generation using different performance envelopes that influence the entire form or only parts of the form according to programmatic demands and/or designer’s preferences. The model generates in every run several design alternatives. The generated alternatives are evaluated in a new type of evaluation model that utilizes numerous fitness criteria, not necessarily used for the generation of the envelopes. These fitness criteria represent another layer of information (besides performance) that the designer receives of the generated form.

In order to select the most fitted solution each criterion is given a normalized priority by the designer.

A total grade is then calculated for each design alternative. Both generation and evaluation process are parametric implying that it is possible to change the number and types of performance envelopes

at any stage of the process and examine the effect of these changes immediately on the design alternative. It also implies that the designer can alter the preference regarding the priorities of the fitness criteria and thus examine changes in the total fitness grade.

The GenPOD model was tested in a case study that examines the applicability of the suggested method in a design of initial form for an office building. The research shows that designing with performance envelopes increases the general performance of the building form by increasing the amount of performance oriented information from which the buildings’ form is generated while allowing to generate an architectural form that embeds a combination of user preference with empiric performance information. Moreover, since the initial form is generated using performance envelopes it adheres by definition to the performances that were used to generate it and does not necessitate an “after the fact” evaluation process, hence, guaranteeing the desired performance of the proposed building.

Performance-oriented design approaches and the use of models such as GenPOD in architectural design are an important step toward a more efficient and sustainable environment.