You are on page 1of 5

Not Just Glass Fins – MIT Brain and Cognitive

Science Project
Neil McClelland, Ove Arup & Partners
155 Avenue of the Americas, New York, NY 10013, USA

Keywords

1=Glass 2=Fins 3=Buckling 4=Seismic

Abstract of insulating glass units typically 3.2m standard solutions for this particular
wide and 2.2m tall with patch fittings configuration, and by the effects of
The MIT Brain and Cognitive Science elevated temperature on the behaviour
Project, Cambridge MA has a large supported off glass fins. Due to the
steps in the elevations, pairs of glass of the interlayer within the fin.
glass enclosed conservatory. Spliced Testing was used to first determine
laminated glass fins support the 13.2m fins at each glass module were adopted
rather than a single glass fin per the safe working load of the fin splices.
high walls. A number of structural After successful completion of these
stability and material technology issues module.
The glass fins span up to 13.2m. tests, a full sized mock-up test was
were encountered and solved. Issues carried out on one module of the
of friction-grip splices in laminated This is too long to manufacture in one
piece so structural splices were required. glazing to structurally proof the overall
glass, effective thickness of laminated design.
glass under high temperatures, Given the glazed skylight strip at the top
of the glazing and the nature of the wall Manufacture of the laminated glass
buckling stability of partial fin restraints fin proved to be problematic. The
and accommodation of seismic glass connections, there would be no
direct vertical load path for the weight design team, glazing contactor and
movements were all addressed. Full the manufacturer both hoped to use
scale mock-up testing was carried out of the wall glass in the event of a glass
fin breakage so laminated glass was a Sentry Glass Plus (SGP) interlayer
to verify structural behavior under for its good performance at elevated
wind and seismic loads as well as to required.
temperatures but time did not allow the

Building Projects
demonstrate redundancy when key The contract form chosen to deliver
the project was performance based effects of the high clamping forces of a

Case Studies
elements were broken. The project friction grip connection to be addressed
was delivered through a collaborative with some latitude for the prospective
contractors to suggest alternate details so pvb interlayers were used.
approach between the architect, facade Installation was generally
consultant, builder, glazing contractor and products. Despite this not being
the most common form of contract straightforward, particularly as the
and glass manufacturer. installation team had used the assembly
within the US market, it was probably
the most cost effective and timely of the mock-up sample as a trial run.
Introduction A few pieces of glass damaged during
format for the client given that the
The MIT Brain and Cognitive Science spliced laminated glass fins required shipping and installation were able to be
Project is located in Cambridge proof of manufacture and performance. replaced without too much difficulty.
Massachusetts on the main MIT campus. Given the size and construction of
Part of the new construction consists the assembly, allowance for seismic Loading and Substructure
of a large glazed conservatory area movement and the required lateral Movements
that is approximately 24m wide, 9m restraints to prevent fin buckling were Wind loading on the vertical walls is
deep and 13m high. The front and one conflicting requirements that needed 2.1 kPa with local high pressure suction
side elevation is stepped and there is carefully designed details. zones on the skylight strips up to 4.9
a 1m wide glazed skylight around the The analysis of the glass fin kPa. The skylights were also subjected
perimeter of the walls. was complicated by both the to snow loads which peaked at 2.4 kPa
The glazed walls are constructed restraint locations as there were no
Figures 1 and 2
The glazed conservatory

GLASS PROCESSING DAYS 2005 - www.gpd.fi 1


where drifting against the main building
may occur.
The building is only in a moderate
seismic zone but as the Conservatory is
so tall, the resulting seismic movements
that the glazing must accommodate are
very high; 125mm in the E-W direction
(in the plane of the main elevation) and
85mm in the N-S direction.
Building movements between the
top and bottom of the wall were up to
40mm with 6mm of that being due to
live load. The dead load was hung from
the top fin connection and all vertical
movements were accommodated within
the fin shoe.

Analysis
The glass fins consisted of two layers of
19mm tempered glass with a 1.52mm
pvb interlayer.
The analysis of the glass fins was
complicated by the use of a pvb
interlayer which will soften at the
internal summertime temperatures
within the Conservatory.
Whilst creep of off-vertical laminated Figure 3
glass at temperatures above 20 °C is Conservatory schematic
understood by many in the industry, and wind loading
temperature effects on laminated
‘window’ glass are usually ignored. This governs. The ‘overestimation’ of Mcr more conservatively, take yh to be zero
is typically not a problem as the mode using the AS1288 approach versus the where it would otherwise be negative in
of failure in ‘window’ glass is governed FEA approach was typically not greater the AS1288 calculations.
by stress where there is usually some than about 1.6 for the conditions Given the doubt of the AS1288 fin
Building Projects

reserve capacity. Where the failure checked. buckling method and the application of
mode is governed by buckling, such as
Case Studies

This is a concern given that the safety a draft code methodology, we carried
in glass beams or fins as we have here, factor against buckling in AS1288 is out our own check of the laminated
there is no reserve capacity and it is 1.7. The effective safety factor against glass fin use FEA buckling analysis. The
unsafe to ignore temperature effects in buckling based upon the FEA results interlayer was directly modelled along
laminated glass. is 1.06. This is not sufficient and with the glass using brick elements.
There is no industry standard we would recommend not using the Temperature properties as supplied
method to calculate effective thickness AS1288 method for calculating Mcr for by the interlayer manufacturer were
for laminated glass at elevated glass fins. used. The results indicated that the
temperatures. The manufacturer chose Whilst a rigorous explanation for fin would not buckle at the design
to use the effective glass thickness this discrepancy has not been derived, load. Interestingly, the Mcr as calculated
calculation method included in prEN it would appear that the case where using AS1288 but taking yh to be zero
13474-1 [1]. The effective fin thickness loading is applied below the centroid was quite close to the buckling load as
was then used in the fin buckling (resulting in a negative yh) does not calculated by the FEA.
equations contained in Appendix H of give the predicted increase in Mcr as per We also calculated Mcr using a
AS 1288:1989 [2]. AS1288. We would recommend that a simpler FEA plate model assuming
We also carried out our own FEA buckling analysis be carried out or, monolithic and layered behaviour as
independent finite element calculations
using brick elements and explicitly
modelling the pvb interlayer properties
at various temperatures. This analysis
was not in good agreement with the
equivalent thickness method and it
indicated a lesser capacity than the
equivalent thickness method. This
discrepancy required some explanation.
To eliminate this potential
discrepancy, a check was carried out on
the fin buckling equation of AS1288.
Finite element analyses (FEA) buckling
checks were carried out on a few fin
geometries.
For the limited sizes of fins checked
here, there is quite a wide discrepancy
between the two approaches with
the FEA bucking method providing
the lower values for the more slender
fins. For the thicker fins, the FEA
predicts much higher Mcr values than Figure 4
the AS1288 approach however, at Comparison of AS1288 and FEA calculated Mcr values on representative fins (300, 400 and 500mm
these values the elastic section capacity deep, 6 to 19 mm thick spanning 6m with a 2 kN/m uniformly distributed load

2 GLASS PROCESSING DAYS 2005 - www.gpd.fi


Figure 6
Fin assembly included 2
friction grip connections
on the long span fins and
2 on the shorter span fins
(shown)

worse than two independent fins. Based plates to prevent the glass from

Building Projects
Figures 5 upon this, a simple calculation method cracking. This design was changed by

Case Studies
FEA buckling analysis including interlayer at el- assuming layered behaviour may provide the contractor to a system that relied
evated temperature results that are not unduly conservative upon stiff plastic rather than aluminium
for laminated glass fins. as a spacer material.
The glazing fittings used were
well as using the calculated effective Detailing proprietary Pilkington Planar fittings.
thickness as per the prEN method. The This system did not require any specific
results for the monolithic case (38mm The performance specification
testing or major modifications for the
thick glass) had about a 10% higher contractually allowed the glazing
project. The use of paired fins offset
Mcr than for a full brick FEA model contractor to modify the design to suit
from the plane of the glass meant that
with 20°C interlayer properties. This their preferred details, construction
there was no lateral support as is usually
can explained by the softer-than-glass methods and to a limited extent,
provided by the silicone joint between
interlayer material being accounted for materials. For the glass fin moment
face and fin glass. The typical Pilkington
in the brick model. Interestingly, the splices, the only practical approach was
detail is designed with a significant
equivalent thickness model predicted a for the manufacturer to develop their
amount of lateral movement and
slightly lower Mcr than the layered model own details. A detail was documented
adjustment. This detail alone would not
(19mm fin taking half the load). It does based upon a friction grip connection
have provided any lateral restraint to the
not make sense that two pieces of glass that relied upon an aluminium plate
fins and so plastic spacers were added
laminated together would perform between the glass layers at the clamp
to the detail. These were not located at

Figures 7 and 8
Typical fin fitting and with
spacers added

GLASS PROCESSING DAYS 2005 - www.gpd.fi 3


all connections but typically at the mid- to accommodate just over 10mm because of its increased stiffness and
height fittings on the face glass to allow of movement. Two possible glass higher ‘softening’ temperature when
for lateral movement of each individual movement scenarios were checked. compared to pvb. We were advised
piece of glass as described later. The fist conservatively assumed that the by the manufacturer that much testing
Insulating glass was used in all face glass panels remained horizontal and was carried out on various assembly
glass locations. The vertical glass had all movements were accommodated combinations to proof the friction
a soft low e coat (Suncool HP) and as by shear of the horizontal joints. The grip splice detail required for the fins.
the horizontal glass was laminated for vertical joints remained basically Unfortunately there was not enough
safety, a hard coat low e (Optifloat static but there would be high shear time to complete all the basic material
K) was used. All glass was heat movements at the crossover junction testing to the satisfaction of the fin and
strengthened or tempered. between the horizontal and vertical SGP manufacturers so a pvb interlayer
Given the height of the assembly, the joints. The second option assumed a was used. Given the relatively low
differential lateral seismic movements of degree of in-plane free body rotation of temperature softening of pvb and the
the structural supports were significant. each panel. Shear movements would lateral buckling behaviour of the fins as
These high movements needed to be be less in the horizontal joints and the described above, thicker 19mm glass
designed for at the detail level but the crossover junction than the horizontal layers were required in the fins verses
whole glass assembly also needed to be movement only model but there would 15mm glass layers required for a SGP
checked to ensure there were adequate be some shear in the vertical joints. The interlayer.
loads paths to hold everything together typical joints were designed as 14mm A number of half sized fins were
under seismic movements. wide with 20mm joints at the corners tested to verify the performance of
One design change suggested by the where tension and compression as well the friction grip moment connection.
glazing contractor was to replace the as shear movements will occur. Under Slippage, displacement, lateral
fin head connections that consisted of some movement scenarios, the Dow displacement and strength criteria were
horizontal angles that captured the full Corning DC 795 structural silicone all met.
top of the fin and were welded back to used in the joints would be strained
the main structure with vertical angles greater than the published movement
that only captured the inner edge of capacity. Given that these strains would
the fins. Whilst this change provided a be in shear rather than in tension-
visually ‘cleaner’ fin head connection, compression and published information
it eliminated the rotational restraint at [3] indicated that actual shear strain
the head. This proved to be problematic capacity could be up to 3 times the
when seismic restraint was being published movement capacity (based
considered. One solution option was upon tension-compression movement),
to mobilize the in-plane stiffness of the Dow Corning were asked to provide a
skylight glazing attached to the ‘free’ project specific warranty allowing shear
Building Projects

edge of the fin however this would have movements above 50% (the published
Case Studies

resulted in increased patch fitting sizes movement of DC795) which they did.
and potentially thicker skylight glass so Structural silicone was used in the
lateral restraint cables were added. As joints for two reasons. The first being
the glazing has 90° corners, there was that the lateral stability of the main
nothing stiff for these lateral restraint south wall fins effectively relies upon
cables to attach to so triangulated steel the face glass plane to transmit the
trusses were added at the corners. small but real lateral loads. This is best
The corner fins were orientated at achieved by utilizing structural silicone.
45° to each of the glass walls. These The second reason structural silicone
were stabilized by horizontal struts was adopted was in the unlikely event
Figure 10
running parallel to the shorter east and of major glass failure, it would provide a
west walls. better resistance to collapse than a low Testing of half-size friction grip moment connec-
tion for the laminated fins
Whilst the overall stability of the modulus high movement silicone.
system was addressed by judiciously
Once the main fin design had been
placed rods and struts, there was still Testing verified, a full size mock-up test was
considerable movement between each
Basic material testing was carried carried out. This consisted of one
piece of glass. With a maximum lateral
out by the glass manufacturer. It full height bay of glazing 13.2m high
design movement of about 125mm over
had been hoped to manufacture and 3.2m wide. The testing covered
the full height of the assembly, each of
the glass fins using a SGP interlayer weatherproofing, structural behaviour
the 6 vertical glass models was required
and seismic movement capacity. Each
of these conditions was met with
Figure 9 measured displacements correlating
Head of corner fins show- well with predicted displacements.
ing stabilizing struts along During proof loading, one of the lights
the short elevation and on the face glass broke. The breakage
stabilizing tie along the
long elevation with trian- occurred about 3 seconds into the
gulated end support 10 second proof loading period. The
broken glass remained in place for the
full duration of the test and was not
considered a failure.
After the proof load test, glass
layers within one fin were progressively
broken and the assembly re-tested to
demonstrate robustness of the design in
the unlikely event of fin breakage. The
mock-up was able to withstand at least
50% of the design load with one layer
of glass in the bottom two fin segments
broken.

4 GLASS PROCESSING DAYS 2005 - www.gpd.fi


Figure 11
Full size mock-up testing

Figure 12
Breaking of fin to test robustness of the design

The mock-up process enabled


the installation team to familiarize
themselves with the erection process.
A method for making positional
adjustments of the fin splices with the
face glass attached was developed on
the mock-up which proved to be useful
on site.

Installation
With the benefit of the lessons
learned during the mock-up assembly,
installation on site proved to be
relatively straightforward. One or two
glass elements had edge chips when
they arrived on site and one segment of
fin was broken after installation when
impacted by the access equipment (the Figure 13

Building Projects
assembly performed as per the mock-up
Accidental breakage of
redundancy tests and did not collapse).

Case Studies
fin on site
These pieces of fin were able to be
replaced with the face glass still in place.

Summary
The glazed conservatory on the MIT
Brain and Cognitive Science Project
presented a number of technical
challenges due to its size. The glass
fins needed to be laminated for
structural redundancy and they needed
to be spliced due to their length.
This required the development of a
high capacity friction grip connection
that did not crush the glass or cause
the interlayer to delaminate. The
behaviour of laminated glass fins
at elevated temperature required
careful consideration. In studying
this behaviour, the applicability of the
AS1288 method of fin buckling analysis
was found to be unconservative in some
cases. The size of the conservatory
meant that careful detailing was
required to allow for seismic movements
and approval from the silicone
manufacturer was obtained to impart
greater than published shear strains
into the sealant. On initial inspection,
the glass assemblies look reasonably
conventional but they are more than
just glass fins.
Specialist structural and façade References
Acknowledgements
consultant: Ove Arup & Partners [1] European Draft Standard prEN 13474-1,
Client: MIT Construction manager: Turner August 1997 Glass in building – Design of
Concept architect: Charles Correa Constructions glass panes – Part 1: General Basis of Design
Associates Glazing Contactor: Karas and Karas [2] Australian Standard AS1288-1989 Glass in
buildings – selection and installation
Local architect: Goody Clancy & Glazing Supplier and Manufacturer: W [3] ASTM C 1401-02 Standard Guide for Structural
Associates & W Glass Systems and Pilkington Planar Silicone Glazing

GLASS PROCESSING DAYS 2005 - www.gpd.fi 5

You might also like