Professional Documents
Culture Documents
Kieran Kelly-Sneed PE
ASI Advanced Structures Inc.,
Los Angeles, California
kieran@asidesign.com
Kieran Kelly-Sneed has several years of experience in modeling and analysis of
cable structures, beginning during his education at Cal Poly, San Luis Obispo. At
ASI, he has been involved in the design of several cable supported glass structures
and of many other custom faade projects. His role extends to the job site where
he has provided extensive support during installation of these and other complex
structures. Recently Kieran spent 8 weeks on site at the San Jose Civic Center to
coordinate tensioning of the Rotunda's cable trusses, a complicated operation due
to the flexible steel ribs that anchor them.
Abstract
While cable-net supported glass walls remain unusual, even exotic design solutions for exterior wall enclosures, their highly
transparent nature makes them an attractive option, particularly for building atria. The behavior of cable-net walls under
lateral loading is conceptually easy to understand, but the large displacements likely under design wind events bring new
challenges to Architects and Engineers not familiar with the design of large displacement structures. The design of cable-net
wall systems involves a great deal of coordination between the designers of the cable-net wall and the designers of the base
building structure (who are often different entities), due to the large loads transferred across the boundaries of the cable-net,
and the large displacements which need be accounted for in the detailing of the interface zones. As complex as the design of
cable-net walls can be, this complexity is repeated in the process of installing the wall. Specialist design-build contractors
are typically engaged to ensure that the installation process is successfully completed.
This paper provides background information on the development of cable-net wall systems and technologies, and descriptive
summaries of the process of designing and constructing cable-net wall systems. Different examples of team make-up, and
the roles and responsibilities of the members of the team are described. The fundamental theory behind the behavior of
planar cable-net wall systems is presented and verified through the application of parametric analysis studies. By way of a
project example, a detailed description is given of the design of The New Beijing Poly Plaza, the worlds largest cable-net
wall recently completed in Beijing.
Architectural Ambitions
Since the dawn of the modernist movement, architects have challenged engineers with an ambition for increased transparency
in building faade systems. As engineering technology has evolved, increasingly lightweight structural support systems have
been developed allowing a range of options for the construction of transparent wall and roof systems. Developments in
structural glass technology have allowed the construction of all-glass solutions for short-span conditions, but for longer
spans a steel backup system is typically required. The conventional solution consists of glazing elements spanning between a
series of parallel lightweight steel trusses. Evolutions in the design of trusses have allowed the use of cable-trusses where the
chords of the truss are replaced by cables, as can be seen in the exterior wall bracing at the Tokyo International Forum
(Rafael Violy Architects P.C., New York & Structural Design Group, Co. Ltd., Tokyo, 1996). Tension is maintained in
each chord for all loading conditions by pre-tensioning the chords against fixed support points or a single internal
compression element. This solution can be considered as an optimization of the underlying beam and slab concept. However
a paradigm shift was achieved with the conception and development of the structural solution for the exterior wall of the
Kempinski Hotel at Munich Airport, (Murphy/Jahn, Chicago & Schlaich, Bergermann und Partner, Stuttgart, 1993). This
planar steel net structure resists lateral loads through catenary action in the cables, completely eliminating the need for any
flexural beam elements behind the glass faade.
Cable-Net Precedents
While conceptually simple, due to their minimal structure and large in-service deflections cable-net curtain wall systems may
still be considered an exotic solution for the support of glass walls. The completion of several major walls around the world,
however, has established a proven track record of an achievable scale and level of transparency. Planar two-way cable
systems support and stabilize glass facades through the resistance to deformation of the two-way pre-tensioned net.
Gravitational loads from the glass elements are carried through the attachment nodes to the vertical cables, and up to a
transfer structure in the base building above. Lateral deformations due to wind and seismic loadings are resisted by the
tendency of each of the horizontal and vertical cables to return to its straight line configuration between supports, while being
subject to a perpendicular force. The flexible nature of a planar cable-net under lateral loading means that the critical design
goal is limiting deflection through adjusting axial stiffness of the cables, and the cable pre-tension. In-service deflection
limits under a 50-year return wind loading condition are typically set in the range of L/40 to L/50 (with L corresponding to
the shortest span passing through the considered point) to protect the integrity of the glass and sealants and to minimize a
perception of movement by the buildings occupants. Oftentimes the controlling design criterion is the later issue concerning
occupant perception. If well designed and detailed, a cable-net wall system may be to sustain the design level wind condition
and beyond without structural or sealant failure. However, if the occupants of the building feel a sense of unease in the
vicinity of the cable-net wall, then the project cannot be deemed a complete success.
Since the construction of the cable-net wall at the Kempinski Hotel in 1993, several significant cable-net wall systems have
been installed around the world. As additional projects have been realized, larger and more complex walls have been
designed and constructed, representing an evolution of the original design concept. Significant milestones in the
development of the cable-net supported wall system include:
The Kempinski Hotel in Munich, Germany, the first application of the planar cable-net concept, is 40m wide by 25m tall and
was designed with the principal cable pre-tension installed in the longer-span horizontal direction. The 22mm diameter
cables are spaced at 1500mm on center, and are pre-tensioned to limit deflections to 900mm.
The AOL Time Warner building on Columbus Circle in New York, USA (Skidmore, Owings & Merrill LLP, James
Carpenter Design Associates & Schlaich, Bergermann und Partner, 2003). This cable-net wall is 46m tall by 25m wide and
both horizontal and vertical cables are 28mm in diameter.
The Beijing New Poly Plaza in Beijing, China (Skidmore, Owings & Merrill LLP, 2006). The cable-net wall is 90m tall by
60m wide and consists of 34mm diameter horizontal cables at 1333mm on center, and 28mm vertical cables at 1500mm on
center. The very large spans of this cable-net wall were achieved by faceting the net by folding it across large diameter
cables (235-275mm). This solution will be described in more detail later in this paper.
Service deflections of a cable-net wall can be significantly reduced by curving the wall in an anticlastic fashion: cables
running in the two principal are curved in opposing directions, creating a stiff grid of intersecting nodes. In this arrangement
lateral deformations due to wind and seismic loadings are resisted with less deflection than for a comparable flat net because
the wall has been built with a partially deflected form. This pre-curved configuration of cables avoids the softest portion of
the non-linear lateral response of a planar cable-net (the undeformed, initial condition). When the anticlastic cable-net is
subjected to lateral loads, only half of the cables resist the loads applied in each direction. The cable-net behaves essentially
as two separate one-way cable-nets, rotated 90 degrees from each other, and each optimized to resist lateral loads in a given
direction. Because of this configuration the support points on each side of the cable-net must be able to resist half the total
applied wind loads on the gross area of the wall, necessitating strong boundary conditions on all four sides. Despite only half
portion of the cables being loaded for a given direction of wind loading, the cables which are mobilized offer such an
increase in stiffness over planar cable-nets, that curved cable-nets can efficiently be designed to deflection limits as high as
L/120. They may also be designed in the range of L/40-L/50 using less pre-tension than a comparable planar cable-net. As
with planar cable-nets occupant comfort and perception are often the determining factors in establishing the deflection limit.
Double Curved cable-net projects include:
The Station Place/Securities & Exchange Commission Headquarters building in Washington, USA (Kevin Roche John
Dinkeloo and Associates LLC & ASI Advanced Structures Inc., 2005). This anticlastic cable-net consists of a wall 27.5m
tall by 19m wide and a roof portion 19m wide by 16m deep with a curved delta truss to transfer loads at the corner where the
two nets meet. All cables are 28mm in diameter and deflections are limited to 250mm.
The Seattle-Tacoma International Airport Central Terminal Expansion in Seattle, USA (Fentress Bradburn Architects Ltd. &
ASI Advanced Structures Inc., 2005). This wall runs 108m in length and spans 16m in height. Both horizontal and vertical
cables are 20mm diameter.
CABLE-NET TECHNOLOGIES
Glass Attachments
Drilled glass bolt fittings or patch plates are typically used to fasten the glass to the cable nodes. These attachments must be
designed to accommodate the environmental loads, in addition to the resulting deflections. For this reason rotational glass
bolts are a common solution, allowing upwards of ten degrees of rotation between the glass and its support point. Patch
plates are a more economical solution because a single plate can take the place of four bolts, and the glass does not need to be
drilled. The plate penetrates the glass joint to clamp multiple pieces of glass to a common node. Deflections and rotations
are accommodated by a neoprene bearing pad between the glass and patch plate and by localized deformation of the glass.
Glass Technologies
Glass can be annealed, heat-strengthened, or fully tempered. Heat strengthening and tempering are treatments in which
annealed glass is heated then rapidly cooled, placing its surface in compression and center in tension. This increases the
strength of the glass as higher applied loads can be resisted before the outer surface fractures. Issues of concern when heat
treating glass include nickel-sulfide inclusion, roller wave, and bow or edge warping. Of these, nickel-sulfide inclusion is the
only problem with structural implications; it can cause spontaneous breakage in fully-tempered glass. Heat-soaking after
tempering will usually cause glass with inclusions to break, so the process is often specified in contract documents to reduce
the likelihood of breakage after installation. Most cable-nets employ laminated glass as a safety measure against breakout.
Broken panels remain in place, held by the polyvinyl butyral (PVB) interlayer. Heat-strengthened and fully tempered glass
sheets have different breakage patterns giving added strength after glass breakage to a laminated panel containing both.
Lead architectural designer of the building in the Schematic and Design Development phases.
Architect of Record: Architect responsible for preparation of the construction documents for the building structure, often the
Design Architect.
Design Engineer:
Lead engineering designer of the building in the Schematic and Design Development phases.
Engineer of Record: Engineer responsible for preparation of the construction documents for the building structure,
(Base Building)
often the Design Engineer.
Cable-Net Designer: Lead Designer of the cable-net wall system. This could be the Design Architect, the Design Engineer
an associated designer, or a design-build contractor.
Engineer of Record: Engineer responsible for preparation of the construction documents for the cable-net structure.
(Cable-Net)
This could be the Engineer of Record for the Base Building, but more commonly is the cable-net
design-build contractor.
As can be seen from the above, there are many possible arrangements of different entities that can make up a team for the
design of a cable-net wall system. Several arrangements that have been employed in the past are highlighted in a following
section.
Any structural engineer can design an open web steel roof truss. But it is not an efficient use of
resources when the truss manufacturer can do the detailed design in a fraction of the time, allowing the
structural engineer to select from a catalogue of pre-designed components.
Proprietary Items:
The tension capacity of a swaged steel cable connection is greatly influenced by the exact production
process specific to each manufacturer. As the structural engineer cannot determine this, the cable
assembly is usually provided engineered pin-to-pin as described above.
Design Expertise:
Certain elements or systems may require very specific knowledge to be safely designed, and may
therefore be beyond the experience of the typical structural engineer. In this case the structural
engineer would wisely defer the design of these elements or systems to a specialist. Note that when
the complicating factor is difficulty of design, the chosen course of action may be to place the design
responsibility on a specialist design-build contractor, or to employ the services of a specialist
engineering sub-consultant to assist the structural engineer in the design.
Contractor Expertise: Certain elements or systems may not be difficult to design, but may require specialized knowledge and
skills to safely construct. In such situations the success of the project hangs directly on the ability of
the Contractor to perform the work correctly. In such situations designating this particular portion of
the work as a Contractor Engineered System is a useful tool to ensure that only suitably skilled entities
are contracted to perform the work.
Due to a combination of the reasons listed above, cable-net wall systems are typically designated as Contractor Engineered
Systems in the building design documents, making the design-build contractor the EOR for the cable-net wall system.
However the extent of early involvement in the design by the design-build contractor, and therefore the extent of
development of the initial design by the Design Architect and Design Engineer, vary greatly depending on the team setup
employed.
Building movement, stiffness, and expansion joints must also be coordinated. Preliminary deflection and stiffness
information must be provided to the cable-net designer for use in their modeling, as support displacement can affect the
walls stiffness. Because the cable-net needs a rigid boundary designers try to avoid placing expansion joints across the wall
opening. However, differential building movement can be accommodated through use of springs or hinged elements. Such
issues must also be discussed at the beginning of design.
As the design of the building and the cable-net proceed the cable-net designer coordinates the location of each cable anchor
and associated reaction forces with the design engineer, who in turn supplies support displacement information based on the
cable loads. Design of the attachment detail is usually coordinated between the two engineers, though the engineering
responsibility typically rests with just one. The cable-net EOR always bears responsibility for the cable and swage, beyond
which point responsibility is split on a case by case basis. Embeds in concrete are most often designed by the building EOR.
Depending on the amount of anticipated support deflection, the cable-net designer may coordinate with the base building
EOR just before beginning cable fabrication. Large building deflections may necessitate ordering cables slightly short to
ensure that architectural tolerances or cable thread travel are not exceeded.
The design wind pressure that needs to be considered when calculating the deflections of the cable-net wall is also a subject
of application of the intent of the code. The design wind event should be that prescribed by the governing code (0.02
probability of being exceeded in any given year per UBC97), and the appropriate wind pressures can be established based on
the location and physical form of the building. Some engineers may determine that a code-based approach for establishing
the design wind pressure is acceptable, however due to the unusually flexible behavior of the walls, wind engineering studies
are often recommended. For very large walls aero-elastic wind tunnel studies that account for the flexibility of the cable-net
system in the wind tunnel model are beneficial. Note that wind engineering study reports will generally present
recommended cladding design wind pressures, and structural design wind pressures. In long-span cable-net structures it is
general practice to design the glass and glass attachments for local effects due to the local application of cladding pressures
on each pane, and to design the system as a whole for the global application of structural pressures (cable forces, and stresses
in the glass due to global deformations). This approach is reinforced by IBC2006 table 1604.3 note f which allows the
calculations of deflections of supporting members to be made on the basis of applied wind pressures of 0.7 times the
component and cladding wind loads. This design philosophy is intended to ensure that each pane can withstand the
application of localized peak gusts pressures, whereas the system as a whole need only be designed to withstand the
maximum pressure likely to occur when averaged over the full area of the wall.
For the design of the components of the system, specialist codes and guidelines must be used for each item. The design of
steel cables is not well covered in the typical steel design codes used in the United States, but is well covered in ASCE-19
Structural Applications for Steel Cables in Buildings published by the American Society of Civil Engineers. The German
standard DIN 18800-1 Structural Steelwork, Design and Construction, also provides design procedures for the design of
steel cables. DIN 18800-1 is commonly the code used in the development of load tables included in the product catalogues
produced by the European cable manufacturers.
Analysis Methods
For the simplest of cable-nets, one-way spanning planar systems, hand calculations can be performed to accurately calculate
the behavior of the net under lateral loads (see section concerning cable-net fundamental theory later in this paper). However
for more complicated conditions such as two-way nets, irregular geometries, curved nets, or nets with non-rigid boundary
conditions, computer analysis modeling of the system is required. Due to the highly non-linear response of cable-net systems
under out-of-plane loading, an analysis program with full geometric non-linearity capabilities is essential. Additionally an
analysis program that has been specifically developed to include a cable element as an inbuilt standard element type is
usually a significant advantage. This allows cables to be modeled as single line elements in the analysis model from pin-topin. Some analysis models require that cables be modeled as a series of individual elements connected together in series,
significantly complicating the analysis model. Programs which have been successfully used by SOM & ASI in the analysis
of cable-net walls include S-Frame written by Softek Services Ltd, and Space Gass written by Integrated Technical Software
Pty Ltd. Other programs include RS-Tab by Ing.-Software Dlubal GmbH and Strand7 by Strand7 Pty Ltd.
It should be noted that due to the non-linear response of cable-net systems, linear combinations of analysis load cases are not
applicable and all individual load cases and load combinations must be analyzed in turn. Therefore analysis of complicated
systems subject to many load cases can result in very long analysis run times. This is especially true when the analysis of the
cable system also includes an element of form-finding.
The goal of the analysis model is to represent the cable-net structure at the end of the completion of construction and then
investigate the effect of additional loads that may be applied to the cable-net during its service life. Generally the initial
shape of the cable-net is a known architectural form, and the Cable-Net Designer establishes the initial pre-tension levels that
are required to give the cable-net sufficient stiffness to meet the deflection criteria when subject to the design lateral load.
This is a relatively simple process when the initial geometry of cable arrangements and the initial state of stress of the cables
are independent variables. This is best visualized by considering a planar rectilinear two-way cable-net (such as the
Kempinski Hotel cable-net described earlier in this paper), after installation of the cables but prior to installation of the
clamps at cable intersections. At this point in time the geometry of the nodes is fixed (including end points and crossing
points of all cables), but as each cable connects directly between support points at opposite sides of the cable-net without
being physically connected to any of the other structural elements, the level of pre-tension in any cable can be varied without
affecting any of the other structural elements. The process of establishing the initial state of the cable-net is more
complicated if the initial geometry and initial stress state are not independent. A simple example of this would be a
anticlastic cable-net. In this case each vertical cable forms a parabolic shape curve in one direction, and each horizontal cable
forms a parabolic shape curved in the opposite direction. The initial tension in the vertical cable and horizontal cables react
against each other at each of the cable crossing points. In this example any change of initial tension in any one cable will
result in force redistribution to possibly all of the other cables, and importantly a change in the initial geometry of the system.
Therefore establishing the initial geometry and initial stress state is a complicated, often time consuming iterative process of
establishing a balanced configuration of initial pre-tension values in the cables, maintaining the desired initial geometry,
while achieving the stiffness required to resist lateral loads. As complicated as this form-finding process is in the analysis
model, the process is repeated at full scale in the construction of the wall. A planar wall allows each cable to be stressed to
its initial pre-tension more or less independently, whereas the complex curved wall will require a complex installation
sequence taking into consideration the effect of stressing one cable on all the other cables.
CONSTRUCTION IMPLEMENTATION
tensioning. The jack is extended, engaging the temporary nut and tightening the cable. The permanent nut is run down snug
against the cables ultimate bearing surface and pressure is released from the jack, allowing it to be disassembled.
Commonly, jacks are installed on multiple cables and pulled simultaneously, linked to a common hydraulic reservoir. In this
manner the actual tensioning can be completed with relative speed.
Installation is completed with affixing or tightening the cable clamps, which are often installed at the same time the cables are
hung, and with placement of the glass. Provided that the cable anchors were properly located and installation correctly
planned, the glass often installs very easily. Here, again, the cable-net designer can do much to ease installation by detailing
adjustable connections between the nodes and the glazing. This serves as further evidence of the advantage of design-build
contractors in completing a successful cable-net project.
Planar cable-net walls derive their out-of-plane stiffness from the geometric non-linearity behavior of a flexible element
spanning between pinned supports. The element has negligible bending stiffness, but its axial stiffness provides resistance to
lateral deformation into an alternate configuration that will necessarily be of a longer length that the original straight line
connection between two points. Therefore, the elastic potential energy stored in the extended elements in the deformed shape
can be equated to the work performed by the applied load displacing from the original to the deformed shape.
Although the cables can deform into any number of configurations due to uneven load conditions, the deformed shape of a
single cable under a uniform distributed lateral load w, can be shown to be a parabola of the form of equation (1). This
differs from the typical hanging chain catenary profile due the fact that the wind load can be assumed to be uniform along
the length that is a horizontal projection of the deformed shape:
y = a x2
(1)
The parabolic displaced shape under a constant projected horizontal load can be verified with the following proof:
TV = w x
(2)
Taking moments around the origin for a free body extending from the origin to an arbitrary point on the curve (figure 14):
x
w x + y TH = w x( x )
2
y=
w 2
x
2 TH
(3)
(4)
As the applied load and the horizontal cable tension is constant, the cable will deflect to a parabolic form, with the parameter:
a=
w
2 TH
(5)
If we consider the maximum allowed ratio of deflection under wind loading to span be a constant, R.
y max =
L
L
when x =
R
2
(6)
L
L
= a
R
2
a=
(7)
4
RL
(8)
The angle of deviation the cable makes at the intersection with its support point is:
TVMAX
TH
= tan 1
(9)
TH =
wR L
8
(10)
wL
2
(11)
4
R
(12)
TVMAX =
Gives:
= tan 1
For any given span to deflection ratio R, the shape of the deflected cable is similar: parabolic with a fixed angle of deviation
at the supports that is independent of the span. For the condition R = 45, as used in the design of the New Beijing Poly Plaza
project described in more detail later in this paper, the angle of deviation at the supports is approximately 5.1 degrees. As the
deflected shape is similar for a given value of R, we know that the ratio of extended length to initial length is constant. The
length of the parabolic shape can be established through the summation of a series of small sections of the parabola.
(ds ) dx
(13)
ds 2 = dx 2 + dy 2
(14)
dy 2
ds
= 1 + 2
dx
dx
(15)
S =2
L/2
0
By Pythagoras rule:
S =2
L/2
0
(1 + 4a 2 x 2 ) dx
(16)
L
sinh 1 (aL )
2 2
S = 2
a L +1 +
4
4
a
(17)
R sinh 1
1 16
S
R
=2
+1 +
2
4 R
16
L
(18)
That average geometrical strain in the cable due to the deformation under lateral load is constant for a given value of R:
G AV =
S
1 = 0.00132 when R = 45
L
(19)
The strain varies with the tension in the cable, from a minimum at mid-span, to a maximum at the supports:
T (x ) =
w 2
(
)
+
w
x
2 a
T (x ) =
w
2a
TAV =
2
L
(1 + 4 a x )
2
(20)
(21)
T ( x) dx
(22)
L/2
TAV
R sinh 1
w R L 1 16
R
=
+1 +
2
2
16
8 4 R
T AV = 1.00132 TH when R = 45
(23)
(24)
The average tension in the cable and the average geometrical strain in the cable due to its deformation under lateral load are
both independently fixed for a design solution that meets the design span to deflection ratio. This apparent incongruity is
explained by the fact that although the average tension in the cable is fixed by the geometry and loading condition, this
tension can be achieved by a combination of initial pre-tension in the cable and additional tension due to elastic strain:
T AV = Ti + G AV A E
(25)
For a given geometry and loading condition, there are a series of compatible pairs of axial stiffnesses and pre-tensions which
will exactly meet the deflection criteria. These range from the impractical: zero pre-tension and large axial stiffness, to the
implausible: pre-tension equal to the average tension in the cable under lateral load and effectively zero axial stiffness. As
the elastic modulus of a given material is fixed, reducing the stiffness requires reducing the cable area. As the maximum
cable tension is defined by the loading condition and geometry, there becomes a point when the cable size is reduced to the
point of not having adequate strength capacity to resist the design load conditions. Knowing the material characteristics of
the proposed cable elements, we can establish the minimum cross-section that can resist the design tension. This will result
in the compatible pair with the highest pre-tension, however there is generally no negative economic effect on the design of
the adjacent supporting structure. Although the pre-tension is high, the design of the supporting structure is typically
controlled by the maximum in service tension and not the pre-tension force. Special conditions such as flexible support
conditions consisting of materials adversely affected by high permanent load conditions (concrete or timber) may warrant the
selection of a compatible pair of higher axial stiffness and lower pre-tension. For R = 45:
TH = 5.625 w L kN
(26)
TVMAX = 0.5 w L kN
(27)
TAV = 5.632 w L kN
(28)
TMAX = 5.647 w L kN
(29)
If we use Pfeifer GmbH 1x61 and 1x91 stainless steel spiral strand cables in the diameter range of 20mm to 36mm (a range
which includes the most common sizes in cable-net wall construction), we can then establish from the manufacturers
published technical data the elastic modulus of the cables and an appropriate design stress limit in the cables. The design
stress limit value does not vary significantly for the range of cable diameters considered:
E = 130 kN .mm 2
Z R ,d
A
= 0.688 kN .mm 2
(30)
(31)
For the above example the minimum cable area shall be:
AMIN =
TMAX
mm 2 = 8.208 w L mm 2
0.688
(32)
(33)
The above results confirm that the minimum required cable area, the compatible pre-tension value, and the maximum cable
tension under design loading conditions are all linearly proportional to both the length of span under consideration and the
uniform lateral load.
As an example, a one-way spanning cable-net consisting of 22mm diameter cables (A = 281mm2) could be designed to span
22.8m when subject to a uniform lateral load of 1.5kN.m-1 per cable. The pre-tension for this solution would be 145kN, and
the maximum design tension in the cables would be 193kN.
To verify the theoretical results of the previous section, a series of parametric studies on the required cable sizes and levels of
pre-tension that would be required to install cable-net systems of varying sizes were performed. Analysis models were
created to study both one-way and two-way spanning cable-net systems. To minimize the number of variables involved the
following parameters were set for each of the models considered:
y
y
y
y
y
The analysis models were built using S-Frame version 6.22 (written by Softek Services Ltd.), a three-dimensional analysis
program capable of accurately modeling the geometric non-linearity of cable elements under lateral loads. All elements were
modeled as S-Frame cable elements, each 1.5m long between nodes. Lateral loads were applied as node loads at 1.5m on
center. This loading condition represents the load transfer to the cables through the glass support fittings.
The results of the one-way cable-net analysis models support the theoretical relationships established in the previous section.
A clear linear relationship is observable in the data points resulting from the analysis models (figures 19 & 20). When
plotted against the theoretical linear relationships established in the previous section, we see that the data points differ from
the theoretical relationship by varying degrees. The reason for this variation is that the cable selection is limited to the 16
discrete cable sizes available in the diameter range from 6-36mm (note that the 75m span analysis model requires the use of a
cable diameter larger that those available an estimated cross sectional area was used for a 40mm diameter in this case
although limitations on the cable end fitting technology may preclude the use of such a cable). The cable-net models were
built using cable lengths that were multiples of the typical glass size, and the cable size selected was the next larger size
available than the minimum cable area theoretically required. Certain data points which are close to the theoretical
relationship represent situations when the required area was just less than an available cable, for other data points the next
largest cable size was significantly larger than the minimum required. All analysis model cable area data points are above the
theoretical relationship. A similar situation can also be observed in the data points relating required pre-tension to span
length. As the modeled cable sizes are all larger than the minimum required, the compatible pre-tension values are all lower
than the theoretical maximum pre-tension values. Given these limitations the correlation between analysis model and theory
is very strong.
A further investigation into the behavior of two-way cable-net configurations has also been performed to investigate the
additional efficiency and span capability that can be achieved by running pre-tensioned cables in two-directions across a
square wall, rather than in one direction alone.
We observe from the analysis results that the relationship between span and required cable area as well as between span and
required pre-tensioning, appear to maintain the linear relationships observed in the one-way theoretical and analytical studies
(figures 21 & 22). The analysis results indicate that the two-way cable-net system is capable of achieving a certain span to
the same deflection criteria and resisting the same lateral loading condition, employing the use of both smaller cables and
lower pre-tension values. This does not however necessarily indicate that the two-way cable-net system is more efficient.
As both analysis models considered a cable spacing of 1.5m, the two-way system uses twice as many cables to support a
square cable-net wall as the one-way cable-net system of equivalent dimensions. To evaluate the relative efficiencies of the
two different cable-net systems, a more telling comparison is between the two-way system and a one-way system with an
equivalent density of cables per unit area of wall supported. Doubling the horizontal cable density in a one-way cable-net
system is equivalent to reducing the applied lateral load per cable by one half. As shown in equations (32) & (33), the effect
of this reduction in lateral load per cable is a 50% reduction in minimum area and maximum pre-tension required.
Plotting the theoretical relationships determined for the one-way cable-net system with a reduced cable spacing of 750mm we
see that the two-way analysis model requires greater cable sizes and pre-tension loads that the one-way system. This
suggests that the two-way system is less efficient than the one-way system of equivalent cable density. This is supported by
an evaluation of the status of each of the cables in the two-way system under the lateral load condition. When the center
cables in each span direction exactly meets the specified defection criteria, the other cables running parallel to these deflect to
lesser degrees. If all the cables are designed to the same criteria, most of the cables are under utilized, reducing the efficiency
of the system. All the cables in the one-way system can be simultaneously fully utilized. If we design each cable in a twoway system with a span to deflection ratio that is compatible with the deflection of the whole system, we can hypothesize that
the efficiency of the one-way system can be equaled. This would require varying cable sizes and levels of pre-tension which
would complicate the detailing and installation, potentially resulting in higher overall costs. While the efficiency of the oneway system is high, the detailing of the boundary conditions is complicated. Two sides of the cable-net deflect through the
know angle of deviation of the parabolic displaced shape, however the other two sides must be detailed to accommodate a
sliding condition caused by the maximum cable-net displacement also occurring close to two sides of the net. This difficult
problem is avoided with a two-way system which has small deflections at all sides of the cable-net.
Architectural Concept
The clients goal is for a new headquarters building that represents the companys disparate subsidiaries as a unified whole.
The program for the building contains a wide range of spaces including office, retail, restaurants and the Poly Museum. The
museum, established by one of the companys subsidiaries, has the unique purpose of repatriating Chinas cultural antiquities
through purchases at international auctions.
The project is prominently located at a major intersection along Beijings second ring road, northeast of the Forbidden City.
The sites primary orientation is northeast towards the intersection and beyond to the clients existing headquarters building.
The triangular form minimizes the perimeter length exposed to the elements, while a series of interior atria provide additional
interior surface area to give office areas maximum access to daylight. The result is a simple L shaped office plan that
cradles a large atrium (figure 23).
The Poly Museum is suspended within a lantern in the main atrium space (figure 24). Its crystalline surface of laminated
patterned glass is pleated to increase its light reflecting/refracting qualities (figure 25). Inside the lantern, exhibit and lease
spaces are enclosed by wood walls which control daylight while common circulation areas occupy the void between the
solids and the glazed perimeter walls.
Secondary sunset and all-day atria cut through the west (figure 26) and south (figure 27) legs of the L to act as daylight
chambers for bringing direct sunlight into the main atrium. The exterior walls of these atria are comprised of minimal glass
membranes, supported by two-way cable-nets in order to maximize visual and solar transparency. The main atriums cablenet is stiffened by two V-cables that are in turn counterweighted and kept in tension by the self-weight of the suspended
museum lantern..
The second area requiring special engineering treatment was the museum occupancy termed the lantern, which protrudes
from the southeast core towards the building atrium. The lantern consists of an eight-story tall (starting at level 2) crossbraced steel frame that cantilevers 24m from the building core. There are no column elements underneath the lantern. The
tip of the cantilevered frame is effectively propped by its connection to the primary diagonal cables which simultaneously
stiffen the cable-net wall (figure 29).
The gravity load bearing elements of the lantern are the southeast building core, and the primary diagonal cables which
transfer gravity loads back to the cores at the top of the building. To provide a redundant gravity load path, the braced frame
of the lantern has been designed to achieve a life-safety performance level when cantilevered from the shear-wall core
without the load supporting benefit of the primary diagonal cables.
Lateral forces in the lantern are resisted by the shear wall core at the south-east side acting as a torsion box. The shear wall
core is torsionally restrained by the ground floor slab at level 1 and by its connection to the main building through the level
12 and higher level diaphragms. The lantern floor diaphragms transfer the lateral force to the core on a level by level basis.
Engineering Experts reviewing the design of the project. In addition to the application of increased load factors, the cable
design forces were based on the internal forces resulting from a higher level wind condition (100-year wind event).
The 50-year and 100-year wind loading conditions were determined through careful wind engineering studies performed by
Beijing University. The wind studies included a traditional rigid model of the building massing placed within a proximity
model, and an aero-elastic wind tunnel study. The aero-elastic study was performed on a flexible model of the northeast
cable-net wall, constructed using wires and a flexible membrane and tuned to simulate the anticipated dynamic response of
the cable-net system. This study allowed the effect of feedback between the dynamic behavior of the cable-net and the wind
forcing function to be considered. This additional study was used to verify and modify where appropriate the results of the
rigid model study.
Analysis and testing shows that the New Beijing Poly cable-net wall behaves very much as conceived. The results from the
static non-linear analysis (geometric non-linearity) clearly show that the strategy of subdividing the wall into facets with
shorter individual spans was successful (figure 30). This strategy allows the overall displacements to meet the L/45
deflection limit between hard boundary conditions while maintaining the economic viability of the project.
The final design solution was achieved with the largest of the four primary cables 275mm in diameter and consisting of a
parallel strand bundle of 199 individual 15.2mm diameter 1x7 strands. The largest cable is pre-tensioned to 17,000kN, and
experiences a maximum in service loading of 18,300kN during a 100 year wind event. Using the faceted design solution, the
typical horizontal and vertical cables are limited in diameter to 34mm and 26mm, pre-tensioned to 210kN and 100kN
respectively. Horizontal and vertical cables are spaced at 1333mm and 1375mm on center respectively (figure 31).
The final solution is shown diagrammatically in figure 22. The decoupling mechanism consists of the equivalent of a pulley
at the lower point of the V cables. As the overall building drifts, one half of the V tries to lengthen and the other half tries
to shorten. By connecting them together using a pulley or equivalent mechanism, the strains are able to offset each other,
without inducing additional load in the cables. A cast steel rocker mechanism was designed to perform the equivalent
function of the pulley. By crossing the cables and connecting to the rocker casting arms, the need to provide curved pulley
surfaces and curved sections of the main cable were eliminated (figure 34). The rocker mechanism solution allows the load
path at the connection to be concentric, and also allows the relative lateral drift of the upper half of the building to be
distributed through the upper portion of the cable-net wall. Small relative movements between adjacent nodes on the main
diagonal cables and the cable-net cables are accommodated using pin-ended tie-rod connections.
To evaluate the effectiveness of the design solution prior to completing in-depth analysis of the system, a physical model of
the rocker mechanism was built along with a model of the link concept for reference comparison. The models were
installed in a pin-connected frame, with soft springs installed in series with the diagonal cables. By racking the frame
backwards and forwards, the relative effectiveness of the two concepts could be visually evaluated. The physical model test
demonstrated significant extension in the springs using the link model and negligible extension in the springs using the
rocker mechanism model, highlighting the ability of this connection to decouple the main cables from the base building
lateral system.
The final design of the rocker mechanism included five large castings per connection. The main cable clevis castings are
approximately 4m in length. The clevis castings are designed to pass through each other to maintain concentric load paths
through the connection (figures 35 & 36).
Summary
At 90m tall by 60m wide, the New Beijing Poly Plaza represents a significant step in the design of cable-net wall systems.
Completed in December 2006, the wall is believed to be four times larger than any cable-net wall system built to date. The
engineering challenges of a cable-net wall of this scale required creative approaches to solving issues that have likely not
been addressed on smaller projects. The design solution for the rocker mechanism is an example of an innovative design
solution that employs conventional technologies to solve a truly unconventional problem. The architectural design team
quickly embraced the rocker mechanisms as the central components of the machine that is the structural system of this
atrium. As keystones through which the lantern and the cable-net support each other in a symbiotic relationship, the rocker
mechanisms are celebrated accordingly. Prominently located in the center of the atrium picture window, they are
approachable from the lantern rooftop caf and are expected to become a focal point in the experience of this building.
Spacing
(mm)
Pre-tension
(MN)
Client:
Main V
cables
235-275
11.4-17.0
Design Architect:
Horizontal
34
1,333
0.18-0.25
Design Engineer:
AOR / EOR:
(base building)
Cable-Net
Contractor:
Yuanda /
ASI Advanced Structures Inc.
Vertical
26
1,375
0.08-0.10