Proxy WikiMain Page | Help | Log in

Digital Detailing: Fall 2009

Class Materials

Readings, to be completed for second class:

Syllabus and Research Requirements

Technical Paper Example

Workshop Materials

Tutorials

Assigments

References for "Computing Performance"

Lecture References:

Course Description

Digital Detailing: Testing & Analysis
Columbia GSAPP Fall 2009
Architecture Technology Elective
Instructors Mark Collins & Toru Hasegawa

The goal of the class is to incubate a series of architectural research proposals on the subject of performance and design. Performance is simply defined as "optimally working within a defined metric" - be it structural, energy, lighting, or other numerical analytic. As sustainability becomes a clear direction for the profession, the articulation of an active environment becomes crucial for all designers. The class will look at several performance modeling tools in a workshop environment to examine the relationships between object and environment, shape and energy, occupation and ambiance.

This research will culminate in a short but well-illustrated technical paper with a clear methodology and goal. The research will be understood to be relatively open-ended, ending with speculation on how it could be taken forward, relevance to the building profession, and possible applicability to future projects/research. The most successful researchers will be encouraged and supported to submit to ACADIA or other similar architectural technology conference.

The course will cover the following topics: 

  1. Performance Analysis, Platforms and Issues
  2. Energy Analysis
  3. Solar Analysis
  4. Acoustic Analysis
  5. Computational Fluid Dynamics (Air Flow)
  6. Finite Element Analysis (Structural)
  7. Rationalization or Geometry for Fabrication


The work of the class will continue to be supplemented by the Avery Fab/Con Lab, which will provide access to fabrication facilities for mock-ups. The GSAPP and instructors will also sponsor the further refinement of proposals for publication or in the context of independent studies.

Research Themes

Performance and Environment

Performance is ultimately a measure of effectiveness - it takes place in a context of work and change. It is an exchange of energy through material - embodied in mass, temperature, gravity, sunlight - as well as the resistance (or absorption) to these forces. This is the "operative" work of form - whether it is inert [static] or responsive [dynamic]. Even inert form must reside in a dynamic environment and must operate within a range of contexts. The interactions of object and environment, or the conceptualization of these as a continuum of systems and exchanges, characterizes the performance of any given material system. A performance-based approach is increasingly an architectural imperative with the emergence of sustainability - architecture is now asked to audit and mediate the energies that go into its construction, maintenance and operation. Needless to say, architecture is now conceived as much in the context of intensive performances such as heat and energy as it is a professional practice of of extensive properties such as lengths and budgets.

In order to focus the collective research on performative qualities, we have identified and invested in platforms that allow for intensive qualities (stress, energy, heat) to be available in the design process. These are performance design platforms, which provide crucial feedback to how form will actually interact with the manifold forces acting upon it. In this paradigm, it is crucial for the designer to qualify the relationship between environment and object and how information is allowed to feed-back or feed-forward.

Computational Models

The last decade has been characterized by an increasing agency within architectural form making to describe and construct incredibly complex and precise structures. This ability is underwritten by computational geometry, which identifies optimal and useful means of describing and working with shapes. From the first descriptions of form in simplistic polygons, to B-spline surfaces and more recently in hybrid structures such as subdivision surfaces, advances in computational geometry have been mapped in 1:1 fashion with advances in architectural form production.

The procedures we use to create forms in software are not simply "tools" but are a procedural language of making. Like any language, including DNA and mathematics, the language of form lends itself to abstraction. Primitives and simple procedures are strung together to create more complex forms. A simple change at the beginning (of a sentence, or a model's history) can lead to dramatically different results. From a given set of potential procedures (extrude, loft, scale, translate, etc...) there is an infinite set of potential results - the "solution space". We know that the designer's work is to explore this space, extracting useful results and discarding bad results.

Parametric modeling and other procedural techniques enable us to explore this solution space with ease, simply by changing variables, moving sliders, etc. But how do we characterize (computationally) "useful" or "bad" results? The goal of introducing performance analysis software is to rigorously define "good" and "bad" and utilize this information in the exploration of the search space. This broadens the definition of "model" to include the systems, dependencies and behaviors that work in partnership with the geometry of construction.

Simulation and Visualization

The vast majority of performance software are geared toward validation - input a model, define initial conditions and generate results. "Model" had been loosely defined to include fairly low fidelity geometry but high specificity material properties. This correlated well to the limited capabilities of the computer to handle truly complex scenarios. Heuristic models could establish, with a reasonable level of accuracy, how buildings or structures would perform in "typical" scenarios [see ESP-r an open source building performance software, or EnergyPlus published by the US Dept. of Energy]. These software were operated mostly through textual input, with minimal user interface elements and scarcely any visualization.

Modern performance tools are still only emerging, the most useful of which espouse a "progressive input" style [software such as Ecotect reflect this] - designers are able to quickly and easily input design decisions which are immediately reflected in visual results. As the design drills down into more specific specifications, the accuracy of results increase. Even early massing and orientation can be evaluated. This feedback allows for new dimensions of information to be present at critical moments. The energy impact of changes are simply made apparent, implicitly becoming part of the value judgments made throughout the design process.

The most mature methodologies are based in the highest levels of feedback between this [potentially] visual information and design evolutions. There are still substantial challenges to evolving this process in both its immediacy and its fidelity - the limits of computational power, different modeling standards and implicit assumptions of popular models all must be considered in a performance-based paradigm. The work of semester will inevitably touch upon these limits - our goal will be to identify and promote the most robust means and methods to "see the unseen" and work meaningfully with this information.

Population and Fitness

As software platform become increasingly sophisticated, capable of sharing a broad range of analytics, the amount of information that is available to the architect increases exponentially. In this condition of excess, concepts of "search" and "fitness" become crucial to continue to iterate and refine design concepts.

The relatively low computational overhead in generating form leads to populations of solutions. As soon as a multiple of variables is being tuned, there implicitly exists a multitude of equally viable options. These are referred to as "local optimums", where acceptable solutions are scattered throughout a solution space. This is a "speciation" of design, where the parallel investigation of a multitude of possibilities yields results beyond a single chosen design, developed in a linear fashion. Computation is crucial to this "parallelism" as the unyielding patience and increasing power of processors is highly optimized for these kind of un-thinking operations. The critical insight of the designer is to structure both the process of parallel generation as well as evaluation to identify value within an otherwise equal field. This is not merely a working method but an entirely new paradigm for design to follow, one that puts populations and search as a primary ground for design to take root.

As a necessary part of the semesters research, we want to put forward the concept of populations as an end-game of any process of making. Each paper will be asked to promote a methodology, rather than a single design. This methodology will be tested across a [limited] population to further explore its applications and determine its robustness.

Retrieved from "http://www.proxyarch.com/wiki/index.php5?title=Digital_Detailing:_Fall_2009"

This page has been accessed 150 times. This page was last modified on 24 July 2010, at 01:47.