Engineering Structures 117 (2016) 575–590

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

The mechanical response of cold bent monolithic glass plates during the
bending process
Kyriaki Corinna Datsiou ⇑, Mauro Overend
Glass and Façade Technology Research Group, Department of Engineering, University of Cambridge, Cambridge CB2 1PZ, UK

a r t i c l e i n f o a b s t r a c t

Article history: Cold bending of glass involves the straining of relatively thin glass components, (typically plates), at
Received 11 September 2015 ambient temperatures, and is a low energy and cost effective manner of creating curvilinear forms
Revised 19 February 2016 required in modern glass applications. Cold bending is also popular because it is thought to eliminate
Accepted 7 March 2016
the optical imperfections in curved glass plates that arise during alternative and more conventional ther-
Available online 28 March 2016
mal bending techniques. Experimental and numerical investigations on the cold bending of monolithic
glass plates into anticlastic shapes are undertaken and described in this paper. The aim is to characterise
Keywords:
the cold bending behaviour during the bending process and to evaluate the surface/optical quality of the
Cold bending
Monolithic curved glass
curved plates. Two distinct phenomena of interest are observed: (i) a change in the deformation mode
Anticlastic surface that under particular boundary and loading conditions lead to snap-through buckling and; (ii) a local
Optical quality instability termed ‘‘cold bending distortion” that appears on curved plates when certain applied displace-
Cold bending distortion ment limits are exceeded. This cold bending distortion is found to occur at stresses significantly below
Plate instability the fracture strength of the glass plate, but the distortions can be sufficiently large to breach optical ser-
Snap-through buckling viceability requirements. An optical quality evaluation procedure for predicting the cold bending
response and the resulting optical quality of monolithic glass plates are provided at the end of this paper.
Ó 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://
creativecommons.org/licenses/by/4.0/).

1. Introduction mould. However, different moulds are required for plates of differ-
ent curvature therefore, this method is neither energy nor cost effi-
The demand for flat glass is high and increasing significantly. cient. Furthermore, the optical quality of the curved glass plate is
Since 2009 float glass production has increased by 5% per annum very sensitive to imperfections in the mould. An alternative tech-
to meet this demand. The building industry has by far the largest nique that also requires the glass to be heated above its transition
share of this market, and accounts for around 80% of all the flat temperature is roller bending which can be performed either hor-
glass produced [1]. This increase is a direct result of the recent izontally or by vertical toughening bending. The former is per-
architectural requirements for additional lightness, transparency formed in a horizontal bending toughener during the toughening
and natural light in new buildings, leading to an increased demand process of glass; wherein adjustable tilting rollers are used to form
for larger glass panels. In addition, architectural trends increasingly the desired shape of the heated glass plate. The glass is then sub-
require the use of glass in curvilinear forms to produce smooth mitted to jets of cold air to create the favourable residual stress
free-form facades [2]. The processes available for producing curved profile of toughened glass. On the other hand, in vertical toughen-
glass can be divided in two categories based on whether heat is ing bending, the glass is lowered into the furnace in a vertical posi-
involved in the process. tion and is pressed onto the mould before being toughened. Roller
Static mould bending and roller bending are the two most com- bending methods therefore have adjustable and re-usable
mon techniques used to create curved glass and involve heating ‘‘moulds”, but the optical quality of the glass plate is affected by
the glass above the transition temperature (550 °C), so that it the straightness of the rollers, their position relative to one another
becomes viscous. In the static mould bending method (also known and more commonly by roller wave distortion.
as sag bending), the desired curvature of the glass is obtained by The optical quality of glass can be assessed qualitatively with
allowing the heated flat plate to sag under its self-weight, onto a the use of a zebra board plate (consisting of black and white
stripes). The waviness of the reflected image on the surface of
the glass plate is used to assess whether the level of distortion is
⇑ Corresponding author. acceptable. However, this method is subjective as it relies on the
E-mail address: [email protected] (K.C. Datsiou). experience of the inspector. More recent quantitative methods

http://dx.doi.org/10.1016/j.engstruct.2016.03.019
0141-0296/Ó 2016 The Authors. Published by Elsevier Ltd.
This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
576 K.C. Datsiou, M. Overend / Engineering Structures 117 (2016) 575–590

include: (i) in-contact with the glass gauges (flat bottom or 3-way- fAN,d, and the design compressive residual surface stresses (negative
r 
contact gauge) that are conveyed along the direction of the distor- sign) induced by the thermal/chemical toughening
RES
;
c M;RES
tion while measuring peak to valley height [3,4]; (ii) non-contact
distortion measuring systems involving the use of computer vision
rapp  cf ;app is the maximum design stress on the surface of the glass
and high-resolution cameras [5]. Recommendations such as those induced by loads imposed on the glass during its service life and; cf,
set in EN 12150-1:2000 [3] are often used to determine whether cb is an appropriate safety factor to account for variability during the

the optical quality of the curved plate is acceptable. These limit cold bending process.
the amplitude of the roller wave distortion in fully toughened glass Recent research [17] has also shown that there is a significant
to 0.5 mm over a length of 300 mm [5]. It is currently, possible to additional contribution to strength in heat treated glass attributa-
manufacture toughened glass with significantly smaller roller ble to crack healing, fHeal, therefore, Eq. (1) becomes:
wave distortion amplitudes. In fact, a limit of 0.25 mm is often pre-
 
scribed for high-end applications. More information on these ther- rRES
f AN;d
c
rapp  cf ;app
mal bending methods can be found in [6–11]. rcb 6 M;RES
þ f Heal ð2Þ
Cold bending is an alternative, and relatively recent, technique cf ;cb
of creating curved glass plates. During this process, the curvature
There is a growing interest in cold bent glass [18–23] and cold
is induced elastically at ambient conditions with a relatively small
bent glass has already been used in real world applications [24–
amount of equipment, thereby, making the process energy efficient
27]. Two of the most notable examples of completed projects is
and also allowing the bending to be executed on site. Cold bent
the 125 m long glass shell of the Strasbourg TGV train station con-
glass can be used to generate either single or doubly curved forms.
sisting of 6 mm heat-strengthened laminated plates and the glass
Single curvature/developable glass surfaces are easier to form, but
roof of the Victoria and Albert museum in London that covers an
they are not as popular in architectural design as doubly curved
area of 370 m2 and consists of cold bent insulated glass units that
glass provide a much larger architectural freedom and can be used
are point fixed. Yet, to-date no guideline is available for its manu-
to create smooth, free form, transparent facades. The glass panels
facture/design process.
of various curvatures that are required in this kind of applications
Cold bending of glass may be an efficient method for creating
can be cold bent in shape without any requirement for moulds,
curved glass surfaces, but the limited research conducted on cold
therefore, minimising their cost and making cold bending an
bent monolithic glass plates to-date indicates that it can result in
attractive method of creating curved glass surfaces.
geometric instabilities [28–30]. Staaks and Eekhout [28,29]
Cold bending involves the application of out-of-plane loads on
reported that the free edges of the glass plate change their shape
the glass surface to create the desired shape of the glass plate,
from straight to curved during the cold bending process. Their
the plate is then restrained in its curved state by means of mechan-
bending process involved forcing two corners of the plate out-of-
ical fixings or structural adhesives. The glass is therefore subjected
plane while the other two were point fixed (Fig. 1a), thereby creat-
to permanent bending stresses throughout its service life. Cold
ing a hyper surface. In particular, two deformation modes were
lamination bending [12,13] is another recent technique used to
reported. In the first deformation mode, both diagonals were
restrain the curved glass plate during the cold bending of lami-
curved and the edges preserved their initial straightness (Fig. 1b).
nated glass and involves: (i) bending the un-bonded unit of glass
However, when the out-of-plane displacement at the two corners
plates and interlayer(s) in the desired shape and; (ii) laminating
exceeded 16 times the thickness of the plate, a change in the defor-
the un-bonded bent unit in an autoclave. In this case, the interlayer
mation mode was observed [29]; the plate buckled as one diagonal
preserves the shape of the glass in place though partially, since ini-
straightens and the edges become curved (Fig. 1c). This phe-
tial springback is expected when the restraints are removed after
nomenon is noteworthy because curved edges could result in diffi-
the lamination [12,13].
culties when fixing the plate to the frame and/or aligning the edges
The relatively low strength of annealed glass makes it inappro-
of adjacent glass plates.
priate for cold bending applications, as the radius of curvature that
A simplified analytical model was also proposed by Eekhout and
can safely be introduced in an annealed glass plate is generally too
Staaks [28,29] to predict this buckling instability. The plate was
large to produce significant curvature in the glass plate. Therefore,
considered as a system of two diagonal strips spanning between
toughened glass in the form of heat treated (heat strengthened or
the corners of the plate and intersecting at the centre of the plate,
fully toughened) or chemically toughened glass is often used in
and four rods, one along each of the four edges of the plate. By forc-
such applications. The maximum curvature that can be achieved
ing two corners out of plane, bending increases in the diagonals
in cold bent glass has thus far been limited by the maximum sur-
while the rods connecting the corners are stretched creating an
face stresses, generated during the cold bending process (rcb), that
additional axial compression in the diagonals. A change in the
can be safely be resisted by the toughened glass panel throughout
deformation mode (instability) occurs when the critical Euler
its service life:
buckling stress is exceeded in one diagonal. However, Eekhout
 rRES

f AN;d
c
rapp  cf ;app and Staaks were unable to obtain good agreement between their
rcb 6 M;RES
ð1Þ simplified analytical model and their numerical results.
cf ;cb This change of curvature has also been described by Galuppi
 rRES
 et al. [30] as snap-through buckling. Snap-through buckling, in
where f AN;d
c is the design strength of toughened glass [14– the case of a plate, is a sudden change of deformation in the direc-
M;RES

16], which is the sum total of the design strength of annealed glass, tion of the load in the central regions of the plate. Their analytical

Fig. 1. Deformation modes during the cold bending process (a) undeformed; (b) mode 1 deformation; (c) mode 2 deformation (adapted from [29]).
 K.C. Datsiou, M. Overend / Engineering Structures 117 (2016) 575–590 577

Fig. 2. Cold bending set-up (a) plan view and instrumentation; (b) experimental set-up.

and numerical investigations involved a plate loaded at its four experimentally and modelled numerically. Details of these meth-
corners; downward loading (negative z, Fig. 2a) was applied at ods and their results are provided in Sections 2 and 3, respectively.
two diagonally opposite corners and upward (positive z, Fig. 2a) A further interpretation of the results follows in Section 4 that
at the other two. Their results show that the hyperbolic paraboloid infers the response of the plate during the bending. Furthermore,
form of the deformed plate changes abruptly and its axes of sym- the validated numerical model is modified in Section 4 to investi-
metry are reduced from two to one. This occurs as one of the diag- gate the influence of geometrical characteristics of the plate; orien-
onals straightens while the curvature of the other continues to tation of the plate and; load locations during the cold bending
increase and the edges display significant curvature. An analytical process. Finally, an optical quality assessment procedure is pro-
model [30,31], validated by finite element modelling, was pro- posed in Appendix A to aid designers/manufacturers in achieving
posed by the same authors using a combination of Mansfield’s cold bent glass with an acceptable optical quality.
inextensional plate theory and beam theory to account for the dou-
ble cylindrical shape of the plate. However, the surface quality
resulting from the cold bending process is not quantified or inves- 2. Method
tigated, their analytical results have yet to be validated experimen-
tally, and the influence of different boundary conditions have not Numerical modelling and experimental testing were under-
been investigated. taken to investigate the stability of monolithic glass plates during
Apart from energy and cost related benefits, it is thought that the cold bending process.
cold bending does not affect the optical quality of the glass from
its flat state because, unlike thermal bending, viscous flow is not 2.1. Experimental investigation
required. However, it was recently reported that optical distortions
[32] and unwanted reflections [33] could appear during the bend- Cold bending tests were conducted to produce anticlastic
ing process thereby bringing into question the optical quality of shapes frequently required in free form façades. Anticlastic shapes
cold bent glass plates. This geometric instability and the resulting can be achieved imposing out-of-plane forces along the four edges
surface quality in monolithic glass plates seem to be overlooked in of the plate. The forces may be placed in a variety of locations e.g.
the limited number of experimental studies on cold bent glass to- continuously along the edges, at intermediate district locations,
date, which have thus far focused on the state of stress of the inter- etc. The simplest configuration was adopted in this study wherein
layer in cold bent laminated glass [34,35]. two corners of the plate (B and D, Fig. 2a) were restrained locally
The current study aims to investigate the mechanical response while the remaining two corners (A and C) were subjected to ver-
of point-supported, monolithic glass plates that are cold bent into tical (out-of-plane) loads that were applied incrementally by
an anticlastic shape, developing the fundamental understanding of means of hydraulic jacks (Fig. 2b) actuated by a single manually-
monolithic cold bent glass that will eventually also provide the operated hydraulic pump. The load was applied at a rate of 1.5–
base for future research on the more complicated subject of cold 2.5 N/s. However, the geometric response of the monolithic plate
bent laminated glass. The objective is to verify the presence of is not deemed to be sensitive to the loading rate. Loading was
instabilities during the cold bending process that may lead to per- applied up to fracture of the glass plate in order to capture the full
manent optical distortions. Three different cases of boundary con- extent of the geometric instabilities. The glass plate consisted of
ditions (at the supported corners of the plate) are tested 1000  1000  5 mm thick fully toughened glass and no roller

Fig. 3. Boundary conditions: (a) Case CS: clamped supports; (b) Case PS: pin supports; (c) Case RS: roller supports.
578 K.C. Datsiou, M. Overend / Engineering Structures 117 (2016) 575–590

Table 1
Translational and rotational restraints at corners B and D for each case of boundary conditions (U = translation; R = rotation).

Case Description Restrained DOF Contact area (mm) Glass dimensions (mm) No. of specimens
CS Clamped supports (fixed in translation and rotation) Ux, Uy, Uz, Rx, Ry, Rz 37.5  37.5 1000  1000  5 3
PS Pin supports (fixed in translation, free in rotation) Ux, Uy, Uz 37.5  37.5 1000  1000  5 3
RS Roller supports (free in rotation and translation) Uz 37.5  37.5 1000  1000  5 3

wave distortion was detected on the surface of the specimens. The based on the tangential stiffness derived from the previous con-
edges were polished prior to toughening in order to reduce the verged point on the load–deflection path and the distance along
possibility of premature failure from edge flaws. the tangent path that is specified by the arc length; the iterative
The following three sets of alternative boundary conditions process of the corrective phase follows afterwards to minimise/
were used at the locally supported corners B and D (Fig. 3a–c): eliminate the drift error that is produced by the prediction phase
clamped supports (CS); pin supports (PS) and; roller supports until the solution converges to the equilibrium path. This proce-
(RS). Table 1 provides an overview of the restrained degrees of dure is automatically implemented in Abaqus with an integrated
freedom for each case. Translational restraints were achieved for algorithm. The analysis was set to stop when the displacement at
Cases CS and PS by clamping the glass between two steel plates the loaded corners exceeded 100 mm. The self-weight of the plate
with a contact area of 37.5 by 37.5 mm (Fig. 3a). Additionally, in was modelled as a gravity load, applied prior to the forced dis-
case PS, rotational freedom was provided by an articulated joint placements at the loaded corners.
in the clamping system (Fig. 3b). Finally, in Case RS, corners B The element type and mesh density were selected on the basis
and D were simply supported on spherical supports that were in of displacement and stress convergence tests in order to provide a
turn attached to a sliding bearing, thereby allowing rotation and good balance between accuracy and computational time. The
in-plane translation (Fig. 3c); even though the translation in the model consists of twenty-node, quadratic, brick elements with
+Z direction of corners B and D was not restrained mechanically, reduced integration properties to prevent shear locking. The glass
lifting of the plate from the spherical supports is not an issue in plate is modelled as 2 elements thick, 80 elements wide and 80 ele-
this instance because the reaction forces are not expected to ments long, giving a total of 12,800 elements (Fig. 4). To simulate
change sign during the testing. Prior to testing, the residual stress the boundary conditions that were tested experimentally, the
profile through the thickness of the fully toughened glass was degrees of freedom indicated in Table 1 were restrained accord-
obtained by means of a scattered light polariscope (SCALP-05, ingly at the 18 solid elements (9 on each layer through the thick-
GlasStress Ltd. [36]). Three spot readings were taken on each glass ness of the plate) within the contact area of the supports. In
plate, coinciding with the location of the strain gauges that were particular, all of the top and bottom surface nodes of the 18 ele-
subsequently attached to the top and bottom surface of the glass ments within the 37.5  37.5 mm contact areas of the supports
plate (Fig. 2a). The mean residual compressive surface stress were restrained for Case CS and PS. Contrary, only the bottom sur-
recorded was rRES = 85 ± 6 MPa. Three nominally identical glass face nodes of the central element (12.5  12.5 mm) of the contact
plates were tested per case of boundary conditions to ensure area (37.5  37.5 mm) of the support were restrained for Case RS,
repeatability of the results. The mean failure load (mean failure simulating the area over the sphere of the spherical support that
corner displacement) recorded at the end of the testing was: (a) was restrained from displacing vertically in the experimental set-
586 ± 37 N/(84 ± 8 mm) for Case CS; (b) 645 ± 11 N/(119 ± 1 mm) up. Forced displacements up to 100 mm were applied at the two
for Case PS and; (c) 701 ± 19 N/(135 ± 2 mm) for Case RS. load points.
Displacements and surface stresses were obtained from LVDT Further numerical analyses beyond the experimental set-ups
displacement transducers and rosette strain gauges, respectively (Table 2a and b) were performed by modifying the validated
(Fig. 2a). The vertical (out-of-plane) profile of the support axis
was acquired by a custom-made surface profilometer that com-
prised a horizontal wire gauge and a vertical LVDT displacement
transducer, to measure the in-plane location, lx,y, and out of plane
deflection, dz, respectively. The surface profilometer was deployed
along the support axis at 100 N load intervals.

2.2. Numerical investigation

Numerical models of the experimental set-ups were con-

structed within Abaqus/CAE 6.12-2, using a static Riks analysis
[37] in order to capture any geometric instability that may occur
during the cold bending process. The Riks method is an incremen-
tal analysis used in non-linear problems. In this method the
applied displacements and corresponding deflections are
unknowns for each increment and the progress of the solution is
evaluated using the arc-length. The initial arc length, Dlin, (set to
0.01), which specifies the percentage of the total displacement that
is applied on the first increment of the analysis, and the total arc
length, lperiod, (set to 1.0) provide the initial displacement propor-
tionality factor, (Dkin = Dlin/lperiod) that determines the first incre-
ment of the analysis. Consequently, the procedure consists of a
prediction and a corrective phase for each increment. In the predic-
tion phase a suitable starting point is computed for each increment Fig. 4. Numerical set-up.
 K.C. Datsiou, M. Overend / Engineering Structures 117 (2016) 575–590 579

Table 2
Numerical models: (a) square plates; (b) rectangular plates.

Boundary conditions Length ratio, LR Thickness, h (mm) No. of load points Orientation of the plate during cold bending No. of FEA models
(a)
Case PS 1.00 5⁄ 2 Horizontal 1
Case RS 1.00 5⁄ 2 Horizontal 1
⁄
Case CS 1.00 2, 3, 4, 5 , 6, 7, 8 2 Horizontal 7
Case CS 1.00 5 1 Horizontal 1
Case CS 1.00 5 2 Vertical, horizontal-reverse loading 2
Case RS 1.00 5 2 Vertical, horizontal-reverse loading 2
Case CS 1.50 2, 3, 4, 5, 6, 7, 8 2 Horizontal 7
Case CS 2.00 2, 3, 4, 5, 6, 7, 8 2 Horizontal 7
Case CS 2.50 2, 3, 4, 5, 6, 7, 8 2 Horizontal 7
Case CS 3.00 2, 3, 4, 5, 6, 7, 8 2 Horizontal 7

Boundary conditions Aspect ratio, AR Thickness, h (mm) No. of load points Orientation of the plate during cold bending No. of FEA models
(b)
Case CS 0.50 2, 3, 4, 5, 6, 7, 8 2 Horizontal 7
Case CS 0.75 2, 3, 4, 5, 6, 7, 8 2 Horizontal 7
Case CS 1.25 2, 3, 4, 5, 6, 7, 8 2 Horizontal 7
Case CS 1.50 2, 3, 4, 5, 6, 7, 8 2 Horizontal 7
Case CS 1.75 2, 3, 4, 5, 6, 7, 8 2 Horizontal 7
Case CS 2.00 2, 3, 4, 5, 6, 7, 8 2 Horizontal 7

⁄ denotes model corresponding to experimental set-up.

numerical model in order to: (i) quantify the influence of geomet- surface stress–deflection–load results. These are described in-
rical characteristics, orientation of the plate and load locations on turn in this section.
the optical quality of the glass plate and; (ii) propose a brief optical The variation of deflection at the centre of the plate (point E,
quality evaluation procedure. Fig. 2a), dE,z, versus the load, P, applied on each of the free corners
is shown in Fig. 5a. Each case of boundary conditions produced dif-
3. Experimental and numerical results and observations ferent initial central deflections caused by the self-weight of the
glass plate and the influence of the specific boundary conditions.
The salient results and observations contain the effect of bound- Fig. 5a also reveals that the boundary conditions have a significant
ary conditions on the load–displacement response and on the final effect on the response of the plate during the bending process.
curved shape of the plate; the validation of numerical results; When in-plane displacement is unrestrained at the supported cor-
changes in relative stiffness; local changes in curvature; and ners (Case RS), the load, P, vs. centre deflection, dE,z, relationship is

(a) (b)

(c)
Fig. 5. (a) Load P vs. centre deflection, dE,z, curves for case CS, PS & RC (exp & num data); (b) close-up of (a) for Case CS & PS and; (c) load P vs. applied displacement, dAC,z,
curves for case CS, PS & RS (numerical data are not shown for clarity).
580 K.C. Datsiou, M. Overend / Engineering Structures 117 (2016) 575–590

Fig. 6. Initial and deformed shape: (a) and (b) for case CS and; (c) and (d) for case RS.

almost linear throughout, but when it is restrained the initial linear almost unchanged throughout the bending process. In particular
relationship between load and centre deflection is followed by an for: (a) Case CS and PS: a reduction in the curvature of the support
end of the proportionality (P = 110 N for Case CS and P = 130 N axis occurs after 150 N while the curvature in the load axis contin-
for Case PS). When this load is exceeded, the centre of the plate ues to increase. The support axis experiences a reduction in abso-
remains approximately stationary until the applied load reaches lute values of deflection increments, between subsequent load
a value of 150 N for both cases (Fig. 5b). As the load increases fur- steps (Fig. 7a and c) whereas the deflection increments of the load
ther, the deflection of the centre of the plate decreases gradually axis are almost constant throughout the testing (Fig. 7b and d).
towards its unloaded position. Furthermore, case CS has a stiffer This indicates that, the support axis becomes stiffer as the load
response compared to the other cases of boundary conditions increases and consequently, substantial deflection is confined to
(Fig. 5c). the load axis (Fig. 7b and d); (b) Case RS: an increase in the curva-
It was also observed that the final curved shape of the plate is a ture of the support axis occurs during the bending process while
function of the boundary conditions (Fig. 6a–d). The surface geom- the curvature of the load axis remains almost constant and rela-
etry acquired during the cold bending process is doubly curved. tively small (Fig. 7e and f). This suggests that the load axis is stiffer
Double curvature, involves bending about two axes and could be in this case.
either synclastic or anticlastic. In synclastic surfaces (e.g. parabo- Perhaps even more significantly, a change of sign of curvature
loid geometry), the principal curvatures are of the same sign i.e. was observed along the central third of the length of the support
both centres of principal curvatures are located on the same side axis for Case CS and PS as the load increases beyond 200 N at each
of the surface. Whereas in anticlastic surfaces (e.g. hyperparaboloid corner; the anticlastic curvature is converted to synclastic locally
geometry) the centres of principal curvatures are on alternate sides so that the support axis takes the shape of a two-trough ripple
of the surface. The curvature for Case RS was anticlastic throughout (Fig. 7a and c). At this load, the curvature of the central regions
the experiment (Fig. 7e and f). This was also observed in the early of the plate becomes synclastic. This ripple observed in cold bent
stages of the bending for the other two cases (Cases CS and PS), glass is henceforth referred to as cold bending distortion and can
however, the anticlastic curvature is disrupted in Cases CS and PS have an undesirable effect on the optical quality of the curved glass
when the applied load, P, at each corner exceeds approximately similar to roller wave distortion in fully toughened glass. Qualita-
250 N (Fig. 7a–d). Some asymmetries are evident in Fig. 7 which tive images showing the optical distortion on the surface of the
can be attributed to minor deviations from symmetry in the initial glass were captured during the experimental testing. This was
experimental set-up and are therefore, considered negligible. achieved by means of a ‘‘zebra” board and the use of black spray
The numerical results (Fig. 5a and b) show adequate agreement paint on the bottom surface of the glass to maximise the clarity
with the experimental data. Differences appear when the numeri- of the reflected image on the surface of the glass and to create a
cal model fails to accurately predict: (a) the post-buckling beha- mirror-like effect. Fig. 8a shows the reflection of the black and
viour of the plate for Case CS, where the numerical model shows white stripes and the edge of the zebra board that reveal the cold
a stiffer response and; (b) the load at which the maximum deflec- bending distortion on the cold bent glass plate. Fig. 8b shows the
tion of the centre of the plate occurs for Case PS. These deviations phenomenon resulting from the numerical model.
between experimental and numerical results can be attributed to Stress and deflection data are used in the following 4 load
small imperfections in the experimental set-up. Indeed, it was ranges to describe the mechanical response with increasing load
found that minor misalignments in the boundary conditions during of a cold bent glass plate for Case CS. Principal strain data obtained
iterations of experimental testing resulted in slightly different from strain gauges showed that the directions of the principal
experimental results but the overall load–displacement responses strains are aligned to the support and the load axes. In particular,
remained largely unchanged. the angle between the support axis and the direction of the maxi-
A change in the deformation mode was also observed for all mum principal stresses, h, is relatively small and ranges between
three cases of boundary conditions as the curvature of one diagonal
4.3° and 1.6° during the testing for all the top and bottom stresses
increases while the curvature of the other decreases or remains of points E and S1 (Fig. 2a), making the in-plane shear strain
 K.C. Datsiou, M. Overend / Engineering Structures 117 (2016) 575–590 581

(a) (b)

(c) (d)

(e) (f)
Fig. 7. Experimental results showing the support and load axis profile at 100 N load increments for: (a) and (b) case CS; (c) and (d) case PS and; (e) and (f) case RS.

between the two axes negligible. Therefore, the strain and stress (3) 150–250 N: As the load increases beyond 150 N, a change in
data that are used below refer to the direct principal values in the deformation mode occurs; the load axis continues to
the direction of the support axis. acquire more curvature, while the curvature of the support
axis diminishes (Fig. 7a and b). Within this load range, the
(1) 0–110 N: During the initial stages of the bending process, the centre of the plate starts to deflect back towards its
central regions of the plate displace away from the initial flat unloaded position (Fig. 5b). This coincides with the accumu-
position creating an anticlastic shape, until the applied load lation of top surface tensile stresses in the central regions of
reaches 110 N (Fig. 5b). Strain gauge data and corresponding the plate, rsd,E, (Fig. 9a–c). However, even after this change
stresses, at the top surface of the plate along the direction of in the deformation mode, top surface compressive stresses
the support axis BD (Fig. 2a), rsd, show that up to this load, still dominate the outer third parts of the support axis
in-plane compressive stresses are present along its length (points S1 and S2 from Fig. 2a) (Fig. 9a and d). The maximum
(Fig. 9a). value of compression (7 MPa) at S1 and S2 is reached at the
(2) 110–150 N: When the load exceeds 110 N the central regions load of 250 N (Fig. 9d).
of the plate cease to deflect vertically and remain relatively (4) above 250 N: At 250 N, the top surface compressive stresses
stationary up to a load of 150 N (Fig. 5b). This is accompa- along the support axis start to decrease, as seen at points S1
nied by a corresponding reduction in the top surface com- and S2 (Fig. 9d). At this load a gradual change of the double
pressive stresses at the centre of the plate (Fig. 9b and c), curvature of the plate from synclastic to anticlastic occurs in
whereas the top surface compression in the outer third the central regions of the plate. This leads to the appearance
regions (point S1 and S2, Fig. 2a) of the support axis of the cold bending distortion that manifests itself as a sinu-
(Fig. 9d) continues to increase. soidal deformation along the support axis BD (Fig. 7a and b).
582 K.C. Datsiou, M. Overend / Engineering Structures 117 (2016) 575–590

Fig. 8. Cold bending distortion: (a) distorted reflection of zebra board on cold bent glass and; (b) cold bending distortion in the numerical model: (i) top view and; (ii) side
view (exaggerated 40 times vertically).

(a) (b)

(c) (d)
Fig. 9. Experimental surface stress results in the direction of the support axis for: (a) the support axis (case CS) in loading steps of 100 N, rsd; (b) the centre of the plate, rsd,E,
vs. the applied load; (c) the centre of the plate, rsd,E, vs. the centre deflection and; (d) point S1 (Fig. 2a), rsd,S1, vs. the applied load.
 K.C. Datsiou, M. Overend / Engineering Structures 117 (2016) 575–590 583

(a) (b)
Fig. 10. (a) Cold bending distortion amplitude with increasing load for Case CS & PS (experimental and numerical data) and; (b) experimental curvature results of the edges of
the glass plate.

The amplitude of the cold bending distortion, Adist, i.e. the Strain profiles at points S1 and E (Fig. 11a and b), reveal that
height difference between its highest and lowest point (peak membrane effects are significant along the support axis while the
to trough), continues to increase with increasing load until load axis is predominantly subjected to bending effects. The axial
the end of the test at 600 N (Fig. 10a). The same applies to force that is developed at points S1 and E during the testing can
the curvature of the edges, jedge, of the plate (BA, BC, DA be determined from the membrane strains at these points
and DC, Fig. 2a). (Fig. 12). The results show that the axial compression initially
increases for both points S1 and E until a load of 100 N. Above this
A response similar to that described above was observed for load the axial force decreases at point E and is finally transformed
Case PS. However, for Case RS, neither cold bending distortion to axial tension (180 N) while the axial compression continues to
nor the corresponding tensile stresses on the top surface of the increase at S1. Eventually, the axial compressive stress at point S1
support axis were observed. The top surface stresses in the central reaches a critical value of 25 N/mm at a load of 250 N (Fig. 12),
regions of the RS plate therefore remained in compression and eventually leads to a local instability (local buckling) and the
throughout the cold bending process (Fig. 9b).

4. Discussion

4.1. Interpretation of the mechanical response of the plate

Two phenomena of interest were observed during the cold

bending process: (a) a change in the deformation mode and; (b)
the cold bending distortion. The cold bending distortion was only
apparent in Cases CS and PS. The response for Case CS can be inter-
preted as follows:
The development of surface tensile strains in the central regions
of the plate (150–250 N, Fig. 9b) leads to an increase in the flexural
stiffness of the central parts of the support axis. Therefore, signifi-
cant deflections are confined to the load axis AC while the support
axis starts to decrease in curvature (change in the deformation
mode, Fig. 7a and b). Fig. 12. Axial force along the support axis for points S1 and E during the testing.

(a) (b)
Fig. 11. Cross-sectional strain for point S1 along: (a) the direction of the support axis and; (b) along the direction of the load axis.
584 K.C. Datsiou, M. Overend / Engineering Structures 117 (2016) 575–590

Table 3
Numerical and experimental data for the stress and the amplitude of the cold bending distortion during the bending process.

Case CS Case PS
P (N) dAC,z,num/exp (mm) Adist,num (mm) rt,num,max (MPa) Adist,exp (mm) P (N) dAC,z,num/exp (mm) Adist,num (mm) rt,num,max (MPa) Adist,exp (mm)
0 0 0.00 0.0 0.00 0 0 0.00 0.0 0.00
100
18 0.00 35.9 0.00 100
18 0.00 40.7 0.00
200
32 0.00 54.9 0.00 200
40 0.00 55.6 0.00
300
47 0.19 70.0 0.04 300
55 0.20 68.2 0.05
400
61 0.39 83.4 0.29 400
69 0.42 79.7 0.35
500
76 0.54 95.8 0.46 500
84 0.52 95.6 0.54
600
90 0.64 107.5 0.50 600
105 0.55 105.7 0.61

corresponding gradual change of the double curvature from It is however, shown (see Section 4.2) that snap-through
anticlastic to synclastic in the central regions of the plate (cold buckling could be triggered under a specific case of boundary
bending distortion, Fig. 7a). The sharp peak of axial tension that and loading conditions. Additionally, the increase in the curva-
is developed at point E (Fig. 12) at the load of 250 N is a manifes- ture of the edges, jedge, (BA, BC, DA and DC, Fig. 2a) also
tation of this instability. occurred gradually without any sudden change in shape
The difference in Case RS is that the supported corners of the (Fig. 10b).
plate (B and D) are free to displace in-plane. The absence of in-
plane reaction forces allows the support axis to acquire signifi-
cantly smaller radii of curvature than those of the previous cases
(CS and PS) (Fig. 7a, c, and e). As the load increases, the curvature 4.2. Parametric analysis
of the support axis increases while the load axis retains a constant
curvature (change in the deformation mode). The cold bending dis- The onset of the cold bending distortion is expected to be a
tortion is not triggered in Case RS when the load is applied in the function of the geometrical characteristics, the orientation of the
same direction as the self-weight. However, as shown in plate and the load locations during the bending process. These
Section 4.2.2, the cold bending distortion in Case RS is sensitive parameters were investigated by modifying the numerical model
to the direction of the load with respect to initial imperfections. to suit (Table 2) and the results are discussed in-turn.
Therefore, the cold bending distortion can be attributed to the
forcing of the plate into a non-developable surface generated by
the plate geometry, the boundary conditions and out-of-plane 4.2.1. Geometrical characteristics of the plate
loads. Further numerical simulations were performed to investigate
The numerical and experimental results showing the progress the influence of the initial geometry of the glass plate (thickness,
of the cold bending distortion during the bending process for Case edge length and aspect ratio) on the development of the cold bend-
CS and PS are summarised in Table 3. The cold bending distortion ing distortion. Forced displacements up to 200 mm were applied at
initially occurs when the applied load P exceeds 200 N at each of the two free corners of the plate for Case CS.
the two free corners. At this load, the maximum tensile stress, rt,- The results show that the cold bending distortion occurs at a
max, on the surface of the glass plate, is 54.9 MPa and 55.6 MPa for smaller values of applied displacement during the cold bending
Case CS and PS, respectively. This is significantly lower than the process for thinner plates (Fig. 13a). However, cold bending of
design strength of fully toughened glass which can be approxi- thicker glass plates leads to larger values of maximum distortion
mately estimated using part of Eq. (1) as follows: amplitude. The recommended limit for roller wave distortion
rRES
85 (0.5 mm) is exceeded when the thickness of the glass pla-
f AN;d
¼ 30
¼ 100:8 MPa te P 4 mm. This may compromise the optical quality of the cold
cM;RES 1:20
bent plate.
where rRES is derived from the SCALP measurements. Square plates with different edge length and a thickness of
Therefore, in both cases the cold bending distortion precedes 5 mm were also investigated. This was done by varying the
the fracture of the glass plate and potentially exceeds a serviceabil- length ratio, which is the ratio of the edge length of the tested
ity limit due to the optical distortions occurring in the curved plate. plate, Lp, over the edge length of the reference plate, Lp,0 = 1000 -
The amplitude of the cold bending distortion, Adist, can indicate mm, LR = Lp/Lp,0 = Lp/1000. Numerical results showed that the
whether distortions are visually acceptable or not. European cold bending of plates with larger length ratio, LR, produces
standard limits the amplitude of roller wave distortion in fully smaller values of amplitude of cold bending distortion for the
toughened glass to no more than 0.5 mm [3]. This limit is exceeded same applied displacement at the free corners (Fig. 13b). The
during the cold bending process at an applied load of: P = 420 N maximum distortion amplitude for plates with an edge
(experimental and numerical data) for Case PS and; P = 480 N length P 2000 mm and a thickness of 5 mm does not exceed
(numerical data) and P = 600 N (experimental data) for Case CS the limit of roller wave distortion (0.5 mm) up to an applied cor-
(Fig. 10a). ner displacement of 200 mm.
Overall, the experimental and numerical results support The change in the deformation mode was also found to be a
Galuppi’s [30,31] and Staaks’s [28,29] observation that one diag- function of the aspect ratio, AR, of the plate (Fig. 14a). It was
onal straightens during the bending while the other becomes found that square plates required the smallest magnitude of
more curved (Fig. 7a–f). However, this change in the deforma- forced displacement at the free corners, ddist, to trigger cold bend-
tion mode occurred gradually with increasing load and no gen- ing distortion (Fig. 14b), and exhibited the smallest distortion
eral limit point buckling (snap-through instability), such as that amplitude among the aspect ratios investigated (0.5 6 AR 6
suggested by Galuppi [30] was observed in these cases (Fig. 5b). 2.00).
 K.C. Datsiou, M. Overend / Engineering Structures 117 (2016) 575–590 585

(a) (b)
Fig. 13. Amplitude of the cold bending distortion for plates of different: (a) thickness and dimensions of 1000  1000 mm and; (b) length ratio, LR, and thickness of 5 mm
(numerical data).

Fig. 14. Influence of aspect ratio for 5 mm thick plates: (a) amplitude of the cold bending distortion and (b) applied displacement at the onset of the cold bending distortion
(numerical data).

4.2.2. Cold bending orientation and load locations the cold bending process, shown in Figs. 7e and f/15e and f for Case
Possible influences of plate orientation with respect to grav- RS and Figs. 7a and b/15g and h for Case CS. The resulting config-
ity and load locations were also investigated numerically. In par- uration is based on the direction of the external cold bending loads
ticular, the orientation of the plate was considered pertinent as with respect to gravity; when the self-weight acts in the same
it changes the influence of the self-weight on the cold bending direction as the external loads, the configuration that is initially
distortion. The plate was modelled for Case CS and RS: (a) hor- formed as a result of the self-weight alone, is maintained when
izontally (Hor), with the corner displacement applied in the the external loads are applied and global instabilities are prevented
same direction as the self-weight (this is identical to the exper- (Fig. 7a,b and e,f); however, snap-through instability occurs when
imental and numerical models described in this paper in Sec- the self-weight and the external loads act in opposite directions as
tions 2 and 3); (b) horizontally (Hor-reverse), with the corner the plate changes suddenly from one configuration to the other
displacement applied in the opposite direction to the self- (Fig. 15e,f and g,h). In both cases where snap-through instability
weight and; (c) vertically (Ver), so that the self-weight acts in is noticed, increasing the applied corner displacement further,
the plane of the glass plate and its influence can be considered eventually leads to cold bending distortion (Fig. 16c) (CS
negligible. and RS).
The results show that initial deflections in the plate’s flatness Table 4 summarises both phenomena (snap-through instability
caused by the self-weight have an important influence in the and cold bending distortion) for the different plate orientations of
mechanical response of the plate. In Hor and Ver testing, no Case CS and RS. It is clear that cold bending distortion always fol-
snap-through buckling occurs (Fig. 7a,b and e,f and 15a,b and c, lows snap through buckling, but that snap-through buckling is not
d). However, in Hor reverse testing the plate exhibits a change of always a precursor of cold bending distortion. Finally, Fig. 17
sign of curvature in both the load and the support axis (Fig. 15e,f shows that snap-through buckling has a negligible effect on the
and g,h). This is observed at a load of 63 N and a corner displace- amplitude of the cold bending distortion.
ment of 10 mm for Case CS, while for Case RS the load and corner The influence of the number of the load points, LP, was also
displacement are 78 N and 15 mm, respectively. Fig. 16a and b investigated numerically for Case CS. An increase in the plate’s
shows the deformed plate shape before and after the snap- stiffness is observed when the number of load points, LP, is
through instability for Case RS. Therefore, snap-though instability decreased to 1 and the number of supported corners is simultane-
is triggered when the out-of-plane loads are applied in a direction ously increased to 3. The onset of the cold bending distortion
opposite to the initial out-of-plane deflection induced by self- occurs at larger applied displacement values during the bending
weight (or other surface imperfections) of the plate. Overall, one process (50 mm of applied displacement in comparison to
of two possible configurations occur during the initial stages of 30 mm) while the value of the distortion amplitude is significantly
586 K.C. Datsiou, M. Overend / Engineering Structures 117 (2016) 575–590

(a) (b)

(c) (d)

(e) (f)

(g) (h)
Fig. 15. Influence of the orientation of the plate on the profiles of the support and the load axes. (a) and (b) Ver testing for Case RS; (c) and (d) Ver testing for case CS; (e) and (f)
Hor-reverse testing for Case RS; (g) and (h) Hor-reverse testing for Case CS (numerical data).
 K.C. Datsiou, M. Overend / Engineering Structures 117 (2016) 575–590 587

Fig. 16. Deformed plate shape for Case RS in Hor-reverse testing: (a) under self weight prior to the cold bending; (b) after snap-through instability occurred at a corner
displacement of 15 mm; (c) after the cold bending distortion occurred at a corner displacement of 63 mm (exaggerated 30 times vertically).

Table 4 of interest, a change in the deformation mode and the cold bending
Snap-through instability and cold bending distortion for different plate orientations distortion, may occur during the cold bending process.
for Case CS and RS (U for occurrence, ✗ for absence).
The change in the deformation mode appears as the curvature of
Case CS Case RS one diagonal continues to increase while, depending on the choice of
Hor Hor- Ver Hor Hor- Ver boundary conditions, the curvature of the other diagonal decreases or
reverse reverse remains constant. When in-plane displacement is restrained at the
Snap-through buckling ✗ U ✗ ✗ U ✗ supported corners (Case CS & PS), the support axis becomes progres-
Cold bending U U U ✗ U ✗ sively stiffer during the change in the deformation mode, due to the
distortion membrane action mobilised by the boundary conditions at the sup-
ported corners. Therefore, significant deflections are confined to the
load axis. This phenomenon occurs at an even earlier stage during
the cold bending process if rotation is also restrained at the supported
corners. However, when in-plane displacement is unconstrained
(Case RS), the stiffness of the load axis increases and does not show
a significant increase in curvature compared to the support axis.
Global instabilities (snap-through buckling) were only observed
when the out-of-plane loads on the two free corners were applied
in a direction opposite to the initial out-of-plane deflection
induced by self-weight (or other surface imperfections) of the
plate. This phenomenon appeared in the early stages of the cold
bending process as an abrupt change of the direction of curvature
in both diagonals and can be considered as a particular case of this
change in the deformation mode.
Fig. 17. Cold bending distortion amplitude for different plate orientations of Case
The second phenomenon of interest during the cold bending pro-
CS. cess is the occurrence of a local instability (buckling). As the applied
displacement at the loaded corners increases beyond the change in
the deformation mode in Case CS and PS, a sinusoidal ripple, termed
cold bending distortion, appears along the length of the support axis.
This phenomenon followed in all the cases where snap-through
instability had occurred i.e. even when in-plane displacement was
allowed at the supported corners (Case RS). Therefore, the cold bend-
ing distortion is a function of boundary conditions and preceding
global instabilities that are influenced by the combination of the
self-weight and the direction of the applied load. This ripple can have
a detrimental effect on the aesthetic quality of the curved plate as it
may cause unwanted optical distortions, thereby triggering a ser-
viceability limit state failure of cold bent glass.
The amplitude of the cold bending distortion can be used to
evaluate the optical quality of the curved glass plate as recom-
Fig. 18. Influence of the number of load points on the amplitude of the cold bending mended in EN 12150-1:2000 [3]. The parametric analysis per-
distortion. formed in this study, revealed that the amplitude of the cold
bending distortion is very sensitive to the choice of boundary con-
reduced and is, in this instance, below the recommended limit for ditions, geometrical characteristics of the plate and load locations.
roller wave distortion (Fig. 18). Larger distortion amplitudes were found in plates with pin sup-
ports (Case PS), larger thickness or larger aspect ratio.
5. Conclusions All of the above parameters should be taken into account during
the design of cold bent glass plates in order to avoid snap-though
The aim of this study was to investigate and characterise the instabilities and to ensure the optical quality of the curved plate.
mechanical response of monolithic glass plates when subjected An optical quality evaluation procedure is proposed in Appendix
to cold bending into an anticlastic shape. Numerical modelling A for the cold bending of glass plates in a horizontal position, with
validated by experimental testing revealed that two phenomena two clamped corner supports (Case CS) and loaded at the
588 K.C. Datsiou, M. Overend / Engineering Structures 117 (2016) 575–590

remaining two free corners in the direction of the self-weight to not a comprehensive design guide for cold bent glass, because
avoid snap-through instabilities. This evaluation procedure can be among other things, the strength of glass is not considered. The
used to determine the applied displacement at which cold bending evaluation procedure therefore provides the serviceability limit
distortion occurs and to quantify its amplitude, thereby providing checks required to achieve an acceptable optical quality of the
a measure of the optical quality of the cold bent glass plate. curved monolithic plate at the end of the cold bending process.
Further research could include the use of thin plate theory as a For brevity, results are only provided for glass plates that are sup-
third means of validating the experimental and numerical results ported at two corners with the boundary conditions of Case CS. The
performed in this study. Finally, safety reasons, often dictate that evaluation procedure is based on a numerical parametric analysis
monolithic glass should not be used for structural applications. that was performed for plates of different geometrical characteris-
Therefore, future investigations should also include cold bent lam- tics. In particular, plates with different thickness, aspect ratio (rect-
inated glass, where the polymer interlayer between the two glass angular plates) or length ratio (square plates of different sizes)
plates is expected to influence the mechanical response during were considered in order to populate three dimensional charts,
cold bending as a function of temperature and strain-rate. from which two dimensional views are taken and shown in
Figs. A2 and A3. The charts are useful for establishing the applied
Acknowledgements corner displacement at which the cold bending distortion occurs
and for determining the amplitude of the cold bending distortion.
The authors gratefully acknowledge financial and technical sup- This evaluation procedure is valid for: (a) square plates with edge
port from Eckersley O’Callaghan, and financial support from the length, L, and thickness, h, ranging between 1000 mm to 3000 mm
Research Fund for Coal and Steel of the European Community and and 2 mm to 8 mm, respectively and; (b) rectangular plates with
the Engineering and Physical Sciences Research Council UK (EPSRC). edge length, L, ranging between 500 mm and 2000 mm while the
other edge length is 1000 mm. Their thickness also ranges between
2 mm and 8 mm.
Appendix A. Procedure for evaluating the optical quality of cold
The proposed procedure for assessing cold bending distortion is
bent monolithic glass plates
as follows (Fig. A1):
An optical quality evaluation procedure for predicting cold
(1) Define the geometrical characteristics of the glass plate.
bending distortion in cold bent monolithic glass plates in anticlas-
These include the thickness of the plate, h, the length ratio,
tic shapes is proposed in this section. This evaluation procedure is

Fig. A1. Optical quality evaluation procedure for cold bent monolithic glass plates.
 K.C. Datsiou, M. Overend / Engineering Structures 117 (2016) 575–590 589

(a) (b)

Fig. A2. Applied displacement at the onset of the cold bending distortion for plates of different thickness and: (a) aspect ratio AR and; (b) length ratio LR (numerical data).

Fig. A3. Cold bending distortion amplitude for plates of different thickness and: (a)–(c) aspect ratio AR and; (d)–(f) length ratio for: (a and d) 50 mm; (b and e) 100 mm and;
(c and f) 150 mm of applied displacement (numerical data).

LR for square plates or the aspect ratio, AR, for rectangular (4a) If ddist > dAC,z the desired geometrical and cold bending
plates. characteristics do not cause cold bending distortion
(2) Define the desired cold bending characteristics i.e. the curva- and the optical quality is therefore unaffected.
ture, j, and consequently the forced displacement, dAC,z, that (4b) If ddist < dAC,z, cold bending distortion occurs and the dis-
is to be applied at the two free corners of the plate to pro- tortion amplitude, Adist, should be quantified in order to
duce the desired curvature. These would typically be deter- determine whether it exceeds an acceptable value of
mined from a finite element analysis of the plate on a case- distortion (e.g. limits set in EN12150-1:2000).
by-case basis or from simple geometrical calculations based
(5) Use the desired forced displacement at the free corners, dAC,z,
on the desired radius of curvature;
in Fig. A3a–f to quantify the distortion amplitude.
(3) Determine the applied displacement that triggers the onset
(6) Determine whether the distortion amplitude exceeds
of the cold bending distortion, ddist (Fig. A2).
0.25 mm or another user-prescribed value.
(4) Establish whether cold bending distortion occurs in the pro-
(6a) If Adist < 0.25 mm then the amplitude is considered accept-
cess of achieving the desired cold bending geometry i.e. if
ably low and its effect on the optical quality is negligible.
ddist < dAC,z (Fig. A2a/b).
590 K.C. Datsiou, M. Overend / Engineering Structures 117 (2016) 575–590

(6b) If 0.50 P Adist P 0.25 mm, the cold bending distortion [12] Galuppi L, Royer-Carfagni G. Rheology of cold-lamination-bending for curved
glazing. Eng Struct 2014;61:140–52.
could be visible depending on the location and the inci-
[13] Galuppi L, Royer-Carfagni G. Cold-lamination-bending of glass: sinusoidal is
dence of reflections on the glass. better than circular. Compos B – Eng 2015;79:285–300.
(6c) If Adist > 0.50 mm, the cold bending distortion will be [14] Overend M. Recent development in design methods for glass structures. Struct
clearly visible and is deemed unacceptable. Eng 2010;88(14):18–26.
[15] Overend M, Zammit K. A computer algorithm for determining the tensile
(7) If Adist calculated from step 6, is considered unacceptable, strength of float glass. Eng Struct 2012;45:68–77.
the geometrical and/or bending parameters should be mod- [16] Haldimann M, Luible A, Overend M. Structural use of glass. International
ified accordingly and the cold bending distortion is re- Association for Bridge and Structural Engineering (IABSE); 2008.
[17] Zaccaria M, Overend M. Thermal healing of realistic flaws. J Mater Civ Eng
assessed from step (3). 2016;28(2):04015127.
[18] Chinzi G. Curved glass surfaces: design, process definition & performances. In:
Appendix B. Supplementary material Proc glass performance days, Finland; 2013. p. 336–44.
[19] Fildhuth T, Knippers J. Double curved glass shells from cold bent glass
laminates. In: Proc glass performance days, Finland; 2011. p. 384–9.
Supplementary data associated with this article can be found, in [20] Dodd G, Thieme S. Comparison of curved glass and cold bent panels. In: Proc
the online version, at http://dx.doi.org/10.1016/j.engstruct.2016. glass performance days, Finland; 2007. p. 83–6.
[21] Saksala M. Extraordinary shapes with ordinary costs. In: Proc glass
03.019. performance days, Finland; 2005. p. 371–3.
[22] Feijen M, Vrouwe I, Thun P. Cold-bent single curved glass; opportunities and
References challenges in freeform facades. In: Proc challenging glass 3 & COST action
TU0905, Delft; 2012. p. 829–36.
[23] Vollers K, Veer F. Usage of cold bent glass panes as an approximation for
[1] NSG Group. Pilkington and the Flat Glass Industry 2010. <www.
double curved surfaces. In: Proc glass processing days, Finland; 2003. p. 173–5.
pilkington.com/pilkington-information/downloads/pilkington+and+the+flat
[24] Manara G, Florian C, Tonon L. Cold twisting in point fixed glass: a case study.
+glass+industry+2010.htm> [last accessed June 2015].
In: Proc XXV A.T.I.V. int conf, Parma; 2010.
[2] Eekhout M. Design, engineering, production & realisation of glass structures for
[25] Eekhout M, Lockefeer W, Staaks D. Application of cold twisted temped glass
‘‘free-form” architecture. CWCT members meeting. Bath, United Kingdom;
panels in double curved architectural design. In: Proc glass performance days,
2004.
Finland; 2007. p. 213–20.
[3] CEN (European Committee for Standardization). Glass in building. Thermally
[26] Eekhout M, Niderehe S. The new, cold bent glass roof of the Victoria & Albert
toughened soda lime silicate safety glass – Part 1: Definition and description.
Museum, London. In: Proc glass performance days, Finland; 2009. p. 408–12.
EN 12150-1:2000; 2000.
[27] Vakar L, Gaal M. Cold bendable laminated glass – new possibilities in design.
[4] ASTM Standard C1651-11. Measurement of roll wave optical distortion in
Struct Eng Int 2004;14:95–7.
heat-treated flat glass1. West Conshohocken (PA): ASTM International; 2011.
[28] van Herwijen E, Staaks D, Eekhout M. Cold bent glass sheets in façade
http://dx.doi.org/10.1520/C1651-11. <http://www.astm.org>.
structures. Struct Eng Int 2004;14(2):98–101.
[5] Abbott A. Roller wave distortion. Definition, causes and a novel approach to
[29] Eekhout M, Staaks D. Cold deformation of glass. In: Proc int symp application
accurate on-line measurement. In: Proc glass processing days, Finland; 2001. p.
of architectural glass ISAAG, Germany; 2004.
226–30.
[30] Galuppi L, Massimiani S, Royer-Carfagni G. Large deformations and snap-
[6] Marradi B. Function follows form. Strategies for integrated design of complex
through instability of cold bent glass. In: Proc challenging glass 4 & COST
glass envelopes. In: Proc challenging glass 3 & COST action TU0905 mid-term
action TU0905 final conf, Switzerland; 2014. p. 681–90.
conf, Croatia; 2013. p. 61–8.
[31] Galuppi L, Massimiani S, Carfagni GR. Buckling phenomena in double curved
[7] Neugebauer J. Applications for curved glass in buildings. J Façade Design Eng
cold-bent glass. Int J Non-Linear Mech 2014;64:70–84.
(JDFE) 2014;2:67–83.
[32] Datsiou K, Overend M. Behaviour of cold bent glass plates during the shaping
[8] Teich M, Kloker S, Baumann H. Curved glass: bending and applications. In: Proc
process. In: Proc engineered transparency int conf, Dusseldorf; 2014. p. 125–34.
engineered transparency conf, Germany; 2014. p. 75–83.
[33] Beer B. Complex geometry facades – introducing a new design concept for
[9] Weber F. Curved glass structures. In: Proc glass performance days, Finland;
cold-bent glass. In: Proc glass performance days, Finland; 2013. p. 516–22.
2009. p. 375–80.
[34] Belis J, Inghelbrecht B, van Impe R, Callewaert D. Cold bending of laminated
[10] Von Starck A, Muhlbauer A, Kramer C. Handbook of thermoprocessing
glass panels. HERON 2007;52(1/2):123–6.
technologies, fundamentals, processes, components, safety. Vulkan-Verlag
[35] Belis J, Inghelbrecht B, van Impe R, Callewaert D. Experimental assessment of
GMbH; 2005. p. 605–12.
cold-bent glass panels. In: Proc glass performance days, Finland; 2007. p. 115–7.
[11] Saksala M. Curved and tempered glass – demand, challenge and opportunity
[36] Scattered Light Polariscope SCALP Instruction Manual, Ver. 5.5, GlasStress Ltd.
for the market driven technology. In: Proc glass performance days, Finland;
[37] Abaqus 6.12 Documentation. Analysis User’s Manual, Simulia; 2012.
2003. p. 63–5.