Professional Documents
Culture Documents
Abstract:
This paper presents an overview of the current body of literature available on reciprocal frame structures, the history of the RF as an architectural typology, and a discussion of the links and disconnects
between research and practice in this area. This information is synthesized by organizing the different morphologies observed into a set of heirarchical lineages or phylogenies based on their historical
context and the technological requirements for their rationalization processes.
1
1.0 Introduction
Reciprocal frames comprise a family of structural systems characterized by the interdependent relations of their constituent parts. The term reciprocal frame was coined by the English designer and
builder Graham Brown in order to describe a structural paradigm that had, until that time, been without a name. Reciprocal frame building types have a long, though somewhat obscure history, having
been developed seemingly in parallel by different cultures in response to the constraints of available
materials, but for the most part abandoned following the introduction of modern structural typologies
and increased availability of building materials through trade and the development of better transportation technologies. As our global resources are subjected to greater pressure, there has been
renewed interest in architectural forms that are highly flexible to available materials. Reciprocal frame
systems are efficient in their use of small pieces of material to span large volumes. This has beneficial implications for construction in that it makes availabel a material set that is otherwise unsuitable
for architectural applications. Hardwoods and lower-quality softwoods that cannot be used in other
framing schemes are ideally suited to RF morphology (Thonnisen & Werenfels 2011). This is perhaps
one reason that we have seen a surge in the popularity of reciprocal frames as a research topic over
the past two decades. Another probable factor is the development of more powerful computing tools
tailored to design work. Even simple RF structures have complex geometries (Larsen 2008), and current work explores multiple-unit RF systems that would have been virtually impossible to model (let
alone analyze) with the tools available 20 years ago.
Responsive Unit-intelligence
Exagerated System
Distributed System
done in designing a tensile structure, a reciprocal frame generalizes the role of its constituent parts,
such that similar parts perform in both tension and compression. This is a strength and a limitation of
the RF typology, as similar units are more or less suited to structural demands of their location and
orientation in the structure.
2.3 Gridshells
Gridshells comprise a structural typology that is most closely allied with reciprocal frames in the interdependency of components. Indeed there is some overlap between the two. When a uniform multiunit reciprocal frame system is designed with maximum engagement length between units, its geometry apptoximates a grid. However, there is an essential distinction between the two in the way that
a designer thinks of the fundamental units of each structural type and their means of construction. A
gridshells base units are typically idealized as continuous members that span from one edge of the
structure to the other, connecting to all crossing members. The finished structure derives its strength
from the flex in the members and this connection network over a (usually) double-curved surface. In
construction, this means that the structure is often laid out flat and lifted or otherwise deformed into its
intended form. This differs from the reciprocal frame in both respects. The units of an RF are idealized
as discreet beams that terminate at some length parameter on neighboring units, and in turn provide
termination points for other neighboring units. In practice, these structures are assembled in their final
form, following a sequenced assembly plan (Gelez et al. 2011).
Continuous Members
3.0 History
The concept of the reciprocal frame can be traced to prehistoric building types, though the archaeological record and observations of surviving traditions. Similar systems appear in archetypes such as
the tepees built by the native peoples of North American plains, the prehistoric Hogan Dwellings
(Larsen 2008) and the yurts and gers of the northeast Asian steppes. These simple structures may
not exactly follow the pattern of a reciprocal frame, but they have in common the design approach of
enclosing space with many, relatively small pieces of structural material, arranged in an interdependent system. This efficiency in design appears to have been driven by factors such as the demands of
a nomadic lifestyle and scarcity of available local building material, conditions often correlated to one
another.
Tepee Frame
Hogan Dwelling
In the historical record, the earliest known description of a reciprocal frame structure comes
from Song dynasty China. A depiction of the Rainbow Bridge by Zhang Zeduan dates the concept
to the twelfth century or earlier (di Carlo 2008). Bridge designs of this type also appear in the sketchbooks of Leonardo da Vinci (1452-1519c.e.), suggesting that the idea may have migrated from Asia to
Europe during the renaissance era. Da Vinci expanded the concept seen in his bridge sketches to depict multi-unit grillage patterns which appear to provide the basis for the concept in western research.
Designs by architects Villard de Honnecourt (ca. 1250c.e.) and Sebastiano Serlio (1537c.e.) describe
planar systems with a similar type of interdependent structure composed of rectangular-profile beams
attached with mortise and tenon or bridle joints (Baverel 2000).
The spiral roof framing techniques used by contemporary Japanese and English architects is attributed to the framing innovations developed by the Buddhist monk Chogen for rebuilding temple roofs
in the late twelfth or early thirteenth century. This type of RF is identical in principle to the type now
associated with the term. It is characterized by sloping members arranged in a fan around a central
polygon and resting upon each other in a closed circuit. This also describes the type of RF structure
patented by English designer Graham Brown in 1987 (Larsen 2008).
Simple RF roof
John Wallis, in his Opera Mathematica (1699) was the first to describe the geometries of planar and
inclined systems. Through physical and mathematical models, he also attempted explain load path
resolution for planar RF structures (Baverel 2000).
The work of Olivier Baverel is perhaps most responsible for the current popularity of the RF typology.
His PhD thesis was an important step in defining these structures in a modern context. Here, the geometry of 3-dimensional RF grids is explored using analytic geometry and later generative processes
using genetic algorithms. Extensive structural analysis was also performed to test the effect of varying
the assemblies parameters. Baverel coined the term nexorade to describe these systems. Derived
from the latin, nexor: link, nexorade apparently plays on architectural terminology to place the RF
in a family of repeating-element assemblies (arcade, colonnade, nexorade). The term is elegant, but
competes with Graham Browns reciprocal frame, creating a somewhat confusing and redundant jargon in the research regarding these structures.
[graphics: baverel photo and diagrams]
Patented framing systems, such as the lamella roof invented by Friedrich Zollinger (1920s) further
confuse the vocabulary. Certain types of lamella vaults operate on the same principles as an RF or
nexorade system. A 2010 publication (Tamke et al. 2010) documents the design of a free-form lamella structure using an agent-based form-finding algorithm. The project produces a novel approach
to rationalizing reciprocal frames, but uses a different terminology.
4.0 Morphology
The anatomy of a reciprocal frame assembly can be described as consisting of one or more fan
units, each composed of a number of beams. Parameters specific to this structure include the number
of beams per fan, axial eccentricity, and engagement length in the assembly (Baverel 2000). Multiple
fan reciprocal frame assemblies offer unique opportunities for decorative patterning in the structural
system of a building. They also present unique challenges in the generation of form. In any system
of more than two RF units, there will be at least one member whose ends are both attached along the
mid-span of adjoining members. In non-planar, assemblies, the arrangement of connections makes
it impossible for the axis of one member to intersect with those to which it must connect. This can be
ameliorated by modifying the members individually, or by applying numerical goal-seeking techniques
to the assembly as a whole. After some study, it appears that the domain of reciprocal frame structures can be broken down into families based on the desired macro-geometry and the appropriate
means of rationalization. One interpretation of the RF taxonomy is presented on page 8 of this document, looking at the field through a geometric and historical frame. The families may look very different if parsed using different parameters (e.g. structural or material.)
4.1 Macro: Surface Geometries
Depending on intended form, aesthetic, and the degree of direct control desired in the design process, different approaches are more or less appropriate to the task at hand. In the case of designing
reciprocal frames, families of geometric forms respond more favorably to certain families of strategies.
As is the case with other problems in architectural form generation and rationalization, these strategies can be divided into two groups: top-down and bottom-up. There is some correlation between
these two strategy groupings and the relative complexity of the system, with some notable exceptions.
4.1.1 Top-Down Approaches
Depending on the complexity of the form desired, the rationalization strategies available will be different. When designing RF geometries for radial vaults, simple geometric constructions allow for
analytic solutions with a relatively low degree of complexity in the resulting assembly (e.g. Zollinger
lamella and rainbow bridge patterns.) Domes are most easily rationalized by supplanting the input
surface with a polyhedron that approximates a sphere, and applying transformations to produce the
RF geometry (Senechal et al. 2011). Computational tools have been developed that allow any surface geometry (within limits) to be rationalized as a reciprocal frame structure (Parigi 2014, Song et
al. 2013). These appear to be built on the geometric relationships proposed by Baverel, improving the
applicability of the theory with new optimization strategies and interfaces that allow end-users greater
control.
that Simple RF systems are subject to infinite load paths (Nelson & Kotulka 2007) and can be defined
as statically determinate or indeterminate. (Gelez et al. 2011). Finite element analysis has been used
to show that RF structures can match traditional structures in strength and rigidity (Garavaglia 2013)
while being less sensitive to settling. The greatest potential of these structures appears to be in their
application as deployable architecture, producing spaceframes with relatively simple assembly compared to conventional multi-layer systems.
6.0 Experiments
The final models developed during this project are most closely related to the Zollinger lamella-type
RF system, in that they create beam elements with rectangular cross-sections and curvature on the
top edge. There are two benefits to this geometry in practice. First, the beams are deeper at the
mid-point, where connections are necessarily out-of-axis. Second, the curvature on beam tops more
accurately approximates the input surface, making it easier to plan an elegant sheathing or paneling
strategy. While their units may require more cutomization in systems with variable curvature, lamellatype reciprocal frames are more flexible to input surface geometry and structural constraints because
of their unit cross section parameters and joinery strategy.
6.1 Tools
The test cases for this project were created using Rhino 5 as a modeling and rendering environment,
and Grasshopper as a parametric design engine. The community of Grasshopper users has been
interested in reciprocal frames for several years, and has produced an array of scripts for creating the
geometry. Some of these examples served as inspiration for this project.
6.2 Strategies
The rationalization approaches explored in this project fall into two groups. Both are top-down processes. The first set uses geodesic polyhedra as reference geometry to approximate domed surfaces. The second employs u,v coordinate mapping to transfer a 2D pattern to a 3D surface. Both are
considered top-down processes. In order to gain some understanding of reciprocal frame, and the
relationship between patterning and geometry, initial experiments looked at geodesic polyhedra as a
base for constructing the system. Later tests adopted the lamella-type construction logic as a means
of rationalizing varied curvature and doubly-curved surfaces. The two surface types chosen were a
catenary-profile barrel vault and a torus patch.
11
6.3 Patterning
Test 0 establishes a simple method for constructing the basic pattern of the reciprocal frame. In this
case of these experiments, the basic unit is a 4-member fan. Lines are defined by midpoints on a
bounding rectangle and tangents on a central circle of variable radius. The base unit is propagated by
mapping to a rectangular grid, which can be re-sized to change the proportions and density of patterning.
Mapped pattern
12
intersect it. As in the Zollinger roof system, the top curvature allows sheathing material to be wrapped
evenly over the frame. Lines are placed at the endpoints of each top curve, normal to the curvature of
the vault surface. These act as place-holders to approximate the beam ends. Connecting these lines
produces a bottom edge and closes the loop of the beam outline. Using unique vectors to define the
beam ends creates closed loops, but the majority are non-planar and therefore cannot be used to
define planar surfaces. The problem is remedied by finding the best-fit plane for the vertices of each
beam outline and projecting the geometry to that plane. Another approach defines a single vector for
each pair of end lines. An advantage of using a best-fit plane is its ambivalence to changing curvature
in the surface.
Non-planar loops
The planar surfaces developed in the previous steps approximate the beam geometry, but without accounting for the presence of neighboring beams. The next process is a trimming operation involving
several steps. Rough surfaces are exruded and intersected with trimming planes to create a new set
of beam geometry that is responsive to neighboring elements in length and angle. The data structure
established through this process will also help to locate joinery detailing later on. The following page
shows examples of the variation generated by this definition.
Capture10Detail.PNG
Capture11Detail.PNG
13
beamDepth = 0.5
beamThickness = 0.1
catenaryLength = 1.25*base
catenaryGravity = x0, y0.25, z1.0
beamDepth = 0.5
beamThickness = 0.1
catenaryLength = 1.50*base
catenaryGravity = x0, y0.25, z1.0
beamDepth = 0.5
beamThickness = 0.1
catenaryLength = 2.0*base
catenaryGravity = x0, y0.25, z1.0
beamDepth = 0.5
beamThickness = 0.1
catenaryLength = 2.0*base
catenaryGravity = x0, y0, z1.0
14
Failures occur when beam length is out of proportion with surface curvature. The struts created are
unrealistic and prone to breaking the trimming
operations in the definition. This limit can be quantified as the ratio of the lengths of geodesic cuves
between beam endpoints and straght lines between
the same sets of points.
15
16
17
References
1. Parigi, Dario, Kirkegaard, Poul Henning. Design and Fabrication of Free-Form Reciprocal
Structures, Nexus Network Journal Vol. 16 pp. 69-78. (Turin: Kim Williams Books, 2014)
2. Garavaglia, E. (2013) Collapse behaviour in reciprocal frame structures. Structural Engineering
and Mechanics, 46, ,533-547,,-.
3. P. Song et al. Reciprocal Frame Structures Made Easy, ACM Trans. Graph., vol.32, no. 4, July
2013, New York, NY
4. B. Senechal, C. Douthe, O. Bavarel. 2011. Analytic Investigations on Elementary Nexorades.
5. S. Gelez, S. Aubry, B. Vaudeville. 2011. Nexorade or Reciprocal Frame System Applied to the De
sign and Construction of a 850 m^2 Archaeological Shelter.
6. Tamke, Martin; Riiber, Jacob; Jungjohann, Hauke. Generated Lamella, ACADIA 10; ISBN 978-1-
4507-3471-4] New York 21-24 October, 2010), pp. 340-347
7. Olga Popovich Larsen, Reciprocal Frame Architecture. (London: Architectural Press, 2008)
8. Di Carlo, B.,The Wooden Roofs of Leonardo and New Structural Research. Nexus Network
Journal, 10, 27-38. (2008)
9. Udo Thonnisen, Nik Werenfels Reciprocal Frames: Teaching Experiences. International
Journal of Space Structures Vol. 26 No. 4 (2011)
10. Olivier Baverel, Nexorades: A Family of Interwoven Space Structures, PhD Thesis, University of
Surrey, December 2000
11. Robert Woodbury, Elememts of Parametric Design (London: Routledge, 2010)
12. Erik Nelson & Brandon Kotulka. Infinite Load Path? (Structure Magazine: Oct. 2007)
Image References
1. Model of a 6-strut RF: http://www.reciproboo.org/#/hexayurt-roof-options/4558840976
2. Pavilion at Rice University: http://www.ricegallery.org/new/exhibition/bambooroof.html
3. Diagram of a fan vault: http://www.lookingatbuildings.org.uk/styles/medieval/roofs-and-vaults/stone-
vaulting/fan-vaults.html
4. Lamella framing: https://www.flickr.com/photos/13274317@N00/2699170812/
5. Tension shell: http://www.autoorb.com/singularitiesof-architecture-textile-structures-/3.bp.blogspot.
com*-JEgXFctNcNg*UPZxAqe2ptI*AAAAAAAAAIg*y9WGG4Kg1Ro*s640*tensile-structure.jpg/
18
19
20
21