Mitchell William. 2001. Roll Over Euclid: How Frank Gehry Designs and Builds. New York: Guggenheim Museum Publications, 2001, pp.352-363
Copyright:
Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online from Scribd
Download as pdf or txt
You are on page 1of 12
Roll Over Euclid: How Frank Gehry Designs and Builds
Williarn J. Mitchell
Some architects are angry, really angry, with Frank Gehry. They see his late work as whimsical
and capricious —a betrayal of the stern Modernist commitment to rational problem solving
and economy of means. To them, he is a seducer of the public, promoter of frivolous fash-
ions, and a corrupting influence on impressionable young designers.
Others are envious. They admire the spatial bravura of works like the Guggenheim Museum
Bilbao (1991-97) and the Walt Disney Concert Hall (1987) in Los Angeles, but dismiss them
as singularities made possible by uniquely indulgent clients and generous budgets. Gehry
seems somehow to have slipped the constraints that bind the average architectural Joe.
But both camps get it wrong. Gehry has, in fact, found a way of designing and building that
is far more in tune with the realities of our digitalizing, globalizing age than are the stale dog-
mas of machine-age Modernism. He has created a powerful new architectural language of
computer-constructed curved surfaces, nonrepeating parts, free-form composition, digital
analysis, and globally distributed CAD/CAM (computer aided design/computer aided manu-
facturing) fabrication.
Euclid Rules
You cannot miss the curved surfaces. They challenge a deeply embedded tradition that goes
all the way back to Euclid (ca. 300 8.C.). Euclid's Elements, as every schoolchild used to
know, shows how to construct infinitely varied geometric compositions from points, straight
lines, and arcs of circles. Parallels, perpendiculars, and congruencies figure prominently.
And strictly speaking, the only instruments you need to explore this rigorously beautiful for-
mal universe are a pencil, a straightedge, dividers, and compasses.
Traditional drafting instruments comprise these simple, ancient tools, plus some more-modern
inventions that save the trouble of explicitly executing the commoner Euclidean constructions.
T-squares and parallel bars allow the ready production of parallels. Triangles facilitate inser-
tion of perpendiculars. Graph paper provides a modular framework of both parallels andperpendiculars. Graduated rulers and protractors
simplify the subdivision of lines and angles.
Tracing paper takes the tedious labor out of repli
cating shapes.
Many common fabrication and assembly tech-
niques —craft-based and machine-based—are
closely related. Saws most readily produce straight
cuts and planar surfaces. Bricks and boards fit
together plane-to-plane and at right angles. Brick
walls have parallel sides and rise in parallel
courses. Rolling and extrusion machinery gener-
ates straight lines, while lathes and other turning.
devices yield arcs and circles. Jigs and templates
make it possible to repeat shapes exactly. The tools
ofthe carpenter and the mason give the drafts-
Paul Rudolph, Yale School of Azt and Architecture man's graphic constructions material form at a
(1959-63), New Haven, Connecticut
larger scale—sometimes very literally, such as the
corrugated concrete surfaces of Paul Rudolph’s
‘Art and Architecture Building (1959-63) at Yale University, from the parallel hatching of his
drawings. Architects tend to draw what they can build, and build what they can draw.
‘Small wonder, then, that classical architectural treatises, from Vitruvius to the texts of the
Ecole des Beaux-Arts, are explicitly grounded in Euclid, Even before they introduce the classical
orders, many of these texts provide geometry lessons cribbed from their great Greek predeces-
sor. And John Dee's preface to the first English translation of Euclid (1570) addresses itself
directly to architects; it has good claim, in fact, to be the first theoretical text on architecture
in English.
Early applications of computer-aided design (CAD) technology in architecture simply (and
rather unreflectively) reified this ancient tradition. CAD systems were mostly employed as
accurate and efficient replacements for traditional drafting instruments in the production of
construction documents. They provided points, straight lines, arcs, and circles as basic
graphic primitives, and their operations for inserting these primitives into drawings were
directly based on fundamental theorems of Euclid. Grids and snap operations were the
electronic equivalent of graph paper; copy operations enabled rapid replication of existing
shapes; and drawing “layers” explicitly harked back to the days of transparent tracing-paper
sheets. By greatly enhancing the efficiency of traditional drafting practices, these systems
further marginalized alternative practices.
354The Countertradition Asserts Itself
There has, however, long been an extra-Euclidean countertradition, Despite its marginality,
it has somehow managed to survive. It shows up briefly but significantly, for example, in the
First Book (Chapter Thirteen) of Andrea Palladio's Four Books of Architecture. After remarking
that Vitruvius gives no guidance on construction of the swelling profile of a column, Palladio
sets forth his own method as follows:
The method | use in making the profile of the swelling is this; | divide the shaft of the
column into three equal parts, and leave the lower part perpendicular; to the side of the
extremity of which | apply the edge of a thin rule, of the same length, or alittle longer than
the column, and bend shat part which reaches from the third part upwards, until the end
touches the point of the diminution of the upper part of the column under the collarino.
| then mark as that curve directs, which gives the column a kind of swelling in the middle,
and makes it project very gracefully.
In other words, Palladio resorted to bending a thin elastic spline to create the curve he
wanted. This curve can be described by a precise formula—as later theoreticians of elastically
deformed structures were to demonstrate—but Palladio did not delve into the mathematics.
He simply employed what was, in effect, an efficient analog computation device to produce
the required graphic output.
Antoni Gaudi made similar use of analog computation in designing the complex curved
vaults of the Church of the Sagrada Familia (1883~1926) in Barcelona. In this case, the
device was not a bent spline but a cable in tension. Hung with weights and supported at
each end, the cable traced out a catenary curve. Inverted, the catenary specified an efficient
and beautiful profile for a vault acting in compression.
Other architects, such as Félix Candela, Eduardo Catalano, and Eladio Dieste have worked
with ruled surfaces. These are generated, in vast variety, by taking pairs of curves in space and
connecting them at regular intervals by straight lines—a process most readily carried out by
constructing wooden or wire models. The most familiar are the Saddle-shaped hyperbolic
paraboloid, as in Catalano’s House (1954;
destroyed 2001) in Raleigh, North Carolina,
and the hyperboloid of revolution commonly
seen in the cooling towers of power stations.
Yet other designers have been fascinated by
minimal surfaces. One way to produce these,
without becoming entangled in some fairly
complex and tedious mathematics, is to dip
a wire frame in soap solution; the resulting
Eduardo Catalano, Catalano House (1954), Raleigh,
North Carolina
355soap films immediately give you what you want. With a bit more labor, you can get the same
results with spandex. There is a conference room at the School of Architecture and Planning,
Massachusetts Institute of Technology (MIT), with interior surfaces that were sculpted in this
way by Frank Stella.
At the dawn ofthe computer graphics era, information technologists quickly realized that all
these sorts of curves and surfaces, and many more, could now be generated by digital rather
than analog means. The idea was to encode the mathematical formulas for various types of
curves in software, let the user input parameters specifying particular instances of these
curves, and then employ display routines to trace out these instances on screens or by means
of printers or plotters, An active research community that focused on digital curved-surface
modeling quickly sprang up in the 1960s and early 19703, and developed a repertoire of highly
specialized concepts and techniques with forbidding-sounding names: triangulated surfaces,
parametric curves, bicubics, Coons patches, Bezier curves
and patches, B-splines, and NURBS (non-uniform rational B-
splines). Curved-surface CAD software based upon these con-
cepts became an essential tool of automobile, aerospace, and
ship design. In the entertainment industry, related software
‘was put to work in production of three-dimensional computer
animations. In these fields, with the aid of ever increasing
computer power and ever more sophisticated display technol-
gy, free-form curved surfaces became as straightforward for
designers to handle as straight lines, planes, circles, cylinders,
and spheres were for architects.
In the early 19905, Gehry's partner James Glymph estab-
lished the connection between the worlds of digitally
designed airplanes, automobiles, and animated-cartoon
dinosaurs and that of architecture. The breakthrough project
was a monumental Fish Sculpture (1989-92) for the Vila
Olimpica on the Barcelona waterfront. Its free-form curved
surfaces were digitally modeled using CATIA, a CAD system
CATIA siructural model and primarily intended for use in aerospace. The digital model
Leones eee Building yas used for design development, in structural analysis and
7 design, and in place of traditional drawings as the primary
repository of construction information. This opened the way
to successful application of digital curved-surface modeling in far larger and more ambi-
tious projects, such as the Guggenheim Museum Bilbao, the Walt Disney Concert Hall,
Experience Music Project (EMP}_(1995-2000) in Seattle, and the Ray and Maria Stata
Center (1998-) at MIT. Gehry embarked on an exploration of a formal universe that was no
less rigorously logical and mathematically elegant than that of ancient Greek geometry, but
which—as a practical matter—had been inaccessible before computer-graphics technologyunlocked it. If you were a Platonist, you could say that it had been lurking out there all the
time, waiting for its cultural moment to arrive. It was curtains for the, well, hegemonic dis-
course of straightedge and compasses. Roll over Euclid. Tell Pythagoras the news.
The Role of Three-Dimensional Digitizing and Rapid Prototyping
It is technically possible to design directly at a curved-surface CAD workstation. Some younger
architects have embraced this approach, and Gehry has flirted with it. The amazingly complex
freestanding conference room—a horse-head shaped structure —at the heart of the DG Bank
Building (1995-2001) in Berlin was done this way. But inserting curved-surface
primitives and tweaking control points on the screen is not
necessarily a very fluid or congenial way to explore design
ideas. And it forces reliance on computer visualization as
the primary means of understanding and evaluating forms
and spaces, which turns out not to be entirely satisfactory.
The alternative is to sketch and sculpt forms using standard
physical media as a first step, then to build a closely corre-
sponding digital model by fitting mathematically defined
curves and surfaces to the initial freehand shapes. This is the
strategy mostly favored by automobile designers; curved sur-
face modelers have not made them give up their felt-tip free-
hand sketches, their full-scale taped profiles, or their carefully
crafted clay models. And itis also the strategy of Gehry, who
greatly values the direct tactility of the physical model and the
speed, freshness, and energy of the freehand gesture.
“The trick in this process is to preserve the essential qualities
of the initial two-dimensional or three-dimensional sketch.
These qualities are easily lost, or subtly damaged, if unsuit-
able graphic primitives or inappropriate approximations are
employed. When Jan Utzon first sketched the saillike roof
forms of the Sydney Opera House (1956-73), for example, he
posed very difficult technical problems for the draftsmen who
were to develop and precisely document the design, the engi-
neers who were to analyze it, and the contractors who were to Construction views, DG Bank
build it. Eventually these probléms were solved by introducing Building (1995-2000), Berlin,
a masterful simplification: the free-form surfaces were approx:
imated by triangular patches from the surfaces of spheres, so that the composition became,
in effect, an assembly of smaller versions of Eero Saarinen’s Kresge Auditorium (1953-55) at
MIT flipped on their sides and joined along their upper edges. It was a brilliant move, but it
357carried a heavy penalty. The constructed building, while beau:
tiful in its own way, is much stiffer and more classically geo-
metric than the version that Utzon had originally imagined.
‘That was before computers and NURBS. Today, with the right
equipment, painstaking care, and craftsmanship, it is usually
possible to produce very close approximations that preserve
the important nuances and subtleties of the original. In
Gehry's office, the process begins with the use of a very accu-
rate three-dimensional digitizer to capture vertex, edge, and
surface coordinates from a large-scale physical model. Using
CATIA, mathematical curves and surfaces are then fitted as
closely as possible to these digitized points. Rapid-prototyping
devices, such as computer-controlled three-dimensional
deposition printers and multi-axis milling machines, are then
used to “build back” physical models for visual inspection
and comparison with the original. The process iterates, with
Jorn Utzon skech and longitudinal adjustments as necessary to the digital model, until the
section, Sydney Opera House design team is satisfied
(957-73)
Computer Power Liberates Form
(One of the important uses of the digital model is to provide input data for analysis software.
‘And it is here that a crucial advantage becomes apparent.
In pre-CAD days, when design calculations were carried out by hand, increases in compu-
tational effort resulting from the introduction of complex shapes made an enormous
practical difference. Anyone could quickly calculate the floor area of a rectangular room, for
example. But it took greater mathematical knowledge, and considerably more work, to do
the same for an irregularly polygonal room—and free-form curved rooms were a nightmare.
‘An architect might be able to draw or physically model an arbitrarily shaped building, but
would run into difficulty trying to calculate areas and volumes for costing and construction-
management purposes, accurately visualize shading and shadows, and carry out structural,
thermal, and acoustic analyses to evaluate performance. Producing accurate area and
material takeoffs, and the necessary engineering analyses, might even turn out to be a
technical impossibility.
Today, however, the availability of inexpensive computer power in huge quantities renders these
differences in computational complexity almost entirely unimportant. Even monstrously compli
cated computations of area and volume are no problem for good geometric-modeling software
running on a fast machine, And engineering-analysis procedures have been made far more flexi
358ble and powerful by eliminating the simplifying
assumptions that had restricted their applica-
bility. Thus, for example, the structural analysis
formulas that | painfully learned in architecture
school were restricted to rectangular beams,
circular columns, semicircular arches and
vaults, and the like, but today’s finite-element
sofware can efficiently and accurately analyze
structures of pretty much any arbitrary shape.
The same goes for ray-tracing and radiosity
software that accurately simulates shades and Eero Saarinen, Kresge Auditorium (1954), MIT,
shadows, for computational fluid-dynamics (Cambridge, Massachusetts
software that traces expected airflow in and
around buildings, for dynamic-energy simulation software that predicts thermal performance,
and for acoustic simulation systems.
From a technical viewpoint, simplicity and regularity hardly matter anymore. If designers want,
to emphasize these qualities, they must now do so on other grounds.
CAD/CAM Fabrication
One of these grounds, of course, has long been that simple, regular construction elements
are easier and cheaper to fabricate than more complex ones. Furthermore, if designers can
restrict themselves to using limited numbers of standardized, mass-produced parts, they can
take advantage of industrial economies of scale, The formal implications of these principles
were explored with exquisite poetry by Ludwig Mies van der Rohe, and rather less poetically
by postwar exponents of industrialized component building in Europe. Throughout the
twentieth century, the straightforward logic of industrial production seemed to provide an
unassailable justification for the spare geometries of architectural Modernism
But the implications of today's digitally controlled machinery differ radically from those of
earlier industrial technology. This is vividly illustrated by the contrast between an old-fashioned
printing press and a desktop laser printer. The prifiting press requires considerable time and
effort to set up for a production run, but it then cranks out identical products in high volume
and at low unit cost; itis a typical instrument of mass production. The laser printer has a
higher unit cost, but itis much more flexible. It can rapidly and automatically be reset for every
page, so it costs nothing extra for each one to differ from the last; this property makes it an
instrument of computer-enabled mass customization. Even where small numbers of identical
copies are required, the laser printer is more economical, since it eliminates the high, fixed
preparation and setup costs of the printing press. itis only with large numbers of copies that
unit costs begin to dominate and economies of scale effectively kick in
358Mass customization is particularly attractive in fabrication of
construction components, since buildings are mostly one-off
rather than mass-market products, and itis often difficult to
get sufficiently long production runs to achieve major
economies of scale. It is also well suited to just-in-time pro-
duction and delivery, and similar up-to-date logistical strate-
gies that take advantage of sophisticated, computerized
management tools. And the very nature of buildings creates
demands —which have too often been suppressed in modern
buildings — for variation of components in response to their
particular contexts. There are, after all, good, practical rea-
sons to vary structural members according to their load con-
ditions, windows according to their orientations, and so on.
‘Computer milling of mold for glass
panels for Condé Nast Cafeteria
(2996-2000), New York.
CAD/CAM machinery now provides the means to mass-
customize construction components. These devices are the
architectural equivalents of laser printers. Just as a laser
printer automatically translates a text file into tangible printed output, so a CAD/CAM fab-
rication machine automatically translates a three-dimensional CAD file into full-scale physi-
cal reality, It accomplishes this by performing the necessary physical operations, at high
speed, under very precise digital control
Application of CAD/CAM technology to structural steelwork fabrication has proven to be
particularly effective. Numerically controlled machines can now shape, cut, and drill steel
sections with great efficiency. This means that steel frames can economically be formed into
complex shapes, and that the resulting complicated joints present litle difficulty. Use of
this technology has been crucial for Gehry in such projects as the Guggenheim Museum
Bilbao, Walt Disney Concert Hall, EMP, and Ray and Maria Stata Center. In all of then, CAD
renderings and construction photographs of the naked steel frames are jaw dropping.
For cutting flat sheet material into arbitrary shapes, CAD/CAM laser cutters, water-jet cutters,
and routers have become commonplace. These are much like old-fashioned pen-plotters,
with a powerful cutting device instead of the pen. Gehry has used them extensively to pro-
duce the irregular glass panels for the curtain wall of the Nationale-Nederlanden (1992-96)
Building in Prague, and the sheet-metal cladding panels for the EMP.
Mutti-axis milling machines extend the idea of computer-controlled cutting from two-dimen-
sional sheets to three-dimensional solids. This technology is extensively used in the automo-
bile industry for full-scale prototyping of metal parts. In architecture, it has the potential to
reinvigorate the tradition of non-planar cut stonework, substituting high-speed, extremely pre-
cise mechanical action for the chisels of masons. It is now being utilized, to great effect, in
the still-ongoing construction of GaudI’s Sagrada Familia. Gehry has employed it to shape the
360cut-limestone exterior of the American Center (1988-94) in
Paris; and in full-scale prototyping of a curved stone wall for
the Walt Disney Concert Hall, before that project was
redesigned with a metal skin.
Three-dimensional milling, and other techniques such as
three-dimensional deposition printing, can also be used to
create concrete formwork, and molds for metal and glass.
Thus, for example, the remarkable curved-glass shapes of the
Condé Nast Cafeteria (1996-2000) in New York were pro-
duced by CAD/CAM fabrication of metal molds, then slump- CATIA model of glass panel for Condé Nast
ing heated glass sheets onto the mold surfaces. Cafeteria (1996-2000).
On-site assembly of complex, CAD/CAM fabricated elements can obviously present greater
challenges than assembly of simpler, more standardized pieces. But here, too, the three-
dimensional CAD model helps. It becomes the source of coordinates that can drive laser-
positioning devices and other electronic construction aids.
It can provide clear, detailed renderings of difficult assem-
blies. And it can even control construction robots that
automatically carry out positioning and placement work.
The vigorous exploration of CAD/CAM fabrication technology
by Gehry, his partners, and some highly skilled collaborators
continues his long-standing interest in innovative uses of
materials and construction techniques—an interest that ear-
lier yielded his corrugated cardboard chairs, his use of chain-
link fencing in provocatively unexpected contexts, and other
such surprises. Like any such exploration, it requires embrac-
ing uncertainties and taking risks; and some innovations have
proven more successful than others. But a growing body of
successfully completed projects—with more in the pipeline —
convincingly demonstrates that CAD/CAM fabrication works.
Within schedule and budget frameworks that don’t need to be
extraordinary, it opens up an exciting new way of conceiving
and making buildings. Glass panel mock up for Condé Nast Cafeteria
{1996-2000}, New York.
Repeating and Non-Repeating Parts
In any building, there is a particular balance between repeating and unique elements, which
gives the architecture much of its character. There is also some sort of balance between fac:
tory work and on-site handiwork. As Gehry has realized, CAD/CAM fabrication can shift theseModel, CATIA model, and finished
structure, Fish Sculpture (2989-92)
2 Vila Olimpica, Barcelona
balances—so, one of the critical design issues is to figure out
where they should be established for a particular project.
In the Barcelona fish sculpture, a simple conceptual division
was created between structure and skin. The steel frame is
all CAD/CAM, but the surface is made from copper strips that
were hand-woven together in situ. At Bilbao, the external
shapes and profiles were precisely specified by the CAD model,
and there is no repetition or symmetry of the large-scale
masses; but the titanium cladding panels are another matter.
These were fabricated at a standard size and shape, and they
were then bent and twisted into place on site. The result is
an energetic and compelling interrelationship of mathemati-
cally smooth curves and surfaces at the macro scale; of dis-
crete, repetitive panel rhythms at an intermediate scale; and of
a micro-scale texture of wrinkles and kinks that resulted natu-
rally from the panel fabrication and placement operations. The
macro-scale curves visually connect to the river and the hills,
the intermediate-scale panels to the fenestration patterns of
nearby buildings, and the irregular reflective surface texture to
the cloudy Basque sky.
The Der Neue Zollhof (1994-99) office buildings in Dussel-
dorf resolve the balances in quite another way. Here, the walls
are fluid curves, but the windows are standardized, mass-
produced rectangular elements. The difference between the
two systems is resolved by an adjustable framing element,
which allows each window to be fitted to its particular, non-
standard context.
‘At EMP, nothing on the exterior repeats. The major masses are free form, and if you look closely
at their sheet-metal cladding panels you can see that these are all different. The surfaces were
subdivided according to a logic that took account of local curvature and the technical proper-
ties of the material — particularly the limits on its capacity to bend. The CAD software performed
the immense amount of intricate computation required to unfold each surface facet onto a
plane. Then CAD/CAM cutting machinery accurately produced each unfolded shape from a flat
sheet, These machine-cut sheets were efficiently transported to the site. Finally, they were bent
back and fixed in place to produce exactly the right nonplanar polygonal facets.
Regular grids and repeating parts have been so fundamental to architectural composition for
so long that most designers find it difficult to imagine life without them. But EMP begins to
show how this is possible.Reimagining Design and Construction
Three-dimensional digital models and CAD/CAM construction techniques not only have
profound intellectual implications for designers; they also shake up the organization and
everyday management of architectural practice. It can no longer be business as usual —the
opportunities and demands fundamentally change.
Before computers, most architectural practices logged the
bulk of their billable hours at the design development and con-
struction docurients phases of projects. This is where they
really made their money. Then, when early CAD systems
appeared, workstations took the place of drafting tables, and
electronic computation replaced technical labor. The effect,
pretty unexcitingly, was to automate the production of docu-
ments that had previously been prepared by hand. But more
advanced CAD/CAM processes, as employed by Gehry, are
far more revolutionary; they begin to eliminate, rather than
automate, traditional construction documentation. CAD/CAM
steel fabrication, for example, can now be a largely paperless
process that relies on transfer of digital files rather than shop
drawings
In the now-fading industrial era, catalogues of manufactured
components and materials have played an indispensable
role in architectural practice. Architects spent a lot of their
time searching, selecting, and procuring, In the context of
CAD/CAM, though, the crucial thing is to know the capabili-
ties and availability of the fabrication facilities offered by
various vendors. Then, it becomes possible to design directly _Saperience Music Project EMP)
for those capabilities—a move that provides a great deal of [1995-2000] model being digitized
7 é ad view ofthe completed building,
creative freedom, and involves the architect far more directly Seate
in fabrication and construction processes.
It is easy to imagine that Gehry's high-flying, high-profile, international practice has few
lessons to offer architects applying less advanced resources to more modest projects. But that
‘would be to miss a vital point. Many of the digital-era strategies and techniques that he is pio-
neering now will be increasingly commonplace in the future, just as innovative uses of steel,
concrete, and glass in early Modern buildings eventually became mainstream. His remarkable
late projects will ultimately be remembered not only for the spatial qualities and cultural reso-
nances that they have achieved, but also for the way in which they have suggested that every-
day architectural practice can be liberated from its increasingly sclerotic conventions.
a