Journal of Non-Crystalline Solids 437 (2016) 72–79

Contents lists available at ScienceDirect

Journal of Non-Crystalline Solids

journal homepage: www.elsevier.com/locate/jnoncrysol

Review

Effect of glass thickness on temperature gradient and stress distribution

during glass tempering
Bao-Wei Fan, Ke-Qian Zhu, Qiang Shi, Tao Sun, Ning-Yi Yuan, Jian-Ning Ding ⁎
Jiangsu Collaborative Innovation Center of Photovoltaic Science and Engineering, Changzhou University, Jiangsu Key Laboratory for Solar Cell Materials and Technology,
Center for Low-Dimensional Materials, Micro-Nano Devices and Systems, Changzhou 213164, Jiangsu, China

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

Article history: The thickness of tempered glass is usually more than 3 mm. To achieve thinner tempered glass, it is necessary to
Received 9 November 2015 clarify the stress change during its quenching process. Glass toughening involves high temperatures, which made
Received in revised form 7 January 2016 the real-time measurement of the temperature distribution, stress distribution, and phase changes occurring
Accepted 15 January 2016
difficult. However, these parameters directly affect the strength of the tempered glass. In this paper, for the
Available online xxxx
purpose of evaluating and optimizing the tempering process, nonlinear finite element analysis method has
Keywords:
been used to simulate the distribution of the temperature gradient, and the final residual stress of the glass sam-
Glass tempering ples with different thickness. The geometry and mathematical model to be used were established, and the
Temperature gradient boundary conditions for the simulations were set on the basis of the actual toughening conditions. It is found
Stress distribution that the thickness has a great influence on the quenching period, temperature gradient and stress fields
Thickness distribution.
© 2016 Published by Elsevier B.V.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
2. Simulation analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
2.1. Finite element temperature model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
2.2. Temperature field initial conditions and convection boundary conditions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2.3. Mathematical model of material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2.4. Thermal boundary model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2.5. Finite element analysis model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3. Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.1. Temperature field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.2. Stress field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4. Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

1. Introduction
Tempered glass exhibits numerous advantages over ordinary glass,
such as high resistance to wind and impact toughness [1], owing
to which the former is employed widely in buildings worldwide.
Building-integrated photovoltaic are important devices with respect to
solar energy utilization [2]. These photovoltaic modules are composed
⁎ Corresponding author.
E-mail address: dingjn@cczu.edu.cn (J.-N. Ding).
primarily of solar cells and tempered glass. By decreasing the glass
thickness, the efficiency of the photovoltaic modules can be increased
and their weight can be decreased [3]. Moreover, thin tempered glass
has been used in electronic flat-panel display devices and other such
devices [4]. The thickness of tempered glass is usually more than
3 mm [5]. Although chemically tempered glass can be formed with a
thickness lower than 1 mm, it has a shorter lifespan [6]. In contrast,
the lifetime of physically tempered glass can exceed 30 years [7].
Tempered glass is a glass that has been subjected to a pre-stressing
force. After the toughening process, a uniform compressive stress is

http://dx.doi.org/10.1016/j.jnoncrysol.2016.01.008
0022-3093/© 2016 Published by Elsevier B.V.
 B.-W. Fan et al. / Journal of Non-Crystalline Solids 437 (2016) 72–79 73

Table 1
The thermoviscoelastic characteristics are given in this table [14,22].

Young's modulus Poisson's ratio Glass material factor Solid-state expansion Liquid expansion
(Pa) (K) (K−1) (K−1)

E0 ν D αg αl
7.1 × 1010 0.22 5.5 × 105 1.12 × 10−5 3.0 × 10−5

generated on the glass surface, while a tensile stress forms internally; 2.1. Finite element temperature model
this improves the bending and impact strength of the glass. The residual
stress determines the strength of the glass. According to the American For a three-dimensional glass sheet, the temperature distribution
standard ASTMC1048, the stress in a tempered glass surface should be over a given area is assumed to be a function of the position and time
greater than 6.9 × 107 Pa [8]. For half-tempered glass, it can be coordinates, that is, θ = f(x, y, z, t), where θ is the temperature; x, y,
2.4–5.2 × 107 Pa [9]. For a given set of tempering conditions, the internal and z are the position coordinates; and t is the time. The differential
stress distribution varies with the glass thickness. Glass toughening equation for heat conduction during the tempering process is as
involves high temperatures. Therefore, the real-time measurement of follows:
the temperature distribution, stress distribution, and phase changes
!
occurring within the glass being tempered is difficult. However, these dθ ∂ θ ∂ θ ∂ θ
2 2 2

parameters directly affect the strength of the tempered glass. ρc ¼k 2 þ 2 þ 2 ð1Þ

dt ∂ x ∂ y ∂ z
Usually, the tempering process parameters are optimized through
trial and error, resulting in a long product development cycle, high
costs, and difficulties in ensuring high product quality [10]. In addition, where ρ is the density of the material; c is the specific heat capacity of
ascertaining the temperature-change history for the cooling process the material (glass); t is the time; and k is the thermal conductivity of
on the basis of experience-based judgment and experimental results the material in the x, y, and z directions W / (m K). Further, the following
is very difficult. However, computer simulations of the glass tempering equation is then satisfied:
process, the distribution of the temperature gradient, and the final  
residual stress can make the process of evaluating and optimizing dθ ∂θ ∂θ ∂θ ∂θ
¼ þk þ þ ð2Þ
the tempering process more effective in terms of both time and dt ∂t ∂x ∂y ∂z
cost. Unfortunately, there have been few theoretical simulation studies
in this area. In 1969, based on results reported by Lee et al., Using Eqs. (1) and (2), the entire glass temperature field can be
Narayanaswamy and Gardon proposed the viscoelastic consolidation expressed as an integral form equivalent to a finite number of units:
theory, which predicted the existence of transient stress tempered
glass and could be used to determine the final stress in such glass " " # #
∂θ
[11–13]. In 2010, Nielsen et al. used the finite element analysis software ∫ vol ρc  δθ þ fvgT  fLgT þ fLgT  δTð½D  fLgT Þ dðvolÞ
∂t
ANSYS to simulate the cooling of tempered glass around a hole in a 10-
¼ ∫ S2 δθq dðS2 Þþ∫ S3 δθλðθB −θÞdðS3 Þ ð3Þ
mm-thick glass sample [14].
In this study, we used the nonlinear finite element analysis software
ABAQUS [15] to simulate the toughening process for glass samples with where the subscript “vol” indicates the unit volume; θB is the air
different thicknesses. The ABAQUS is widely used mainly because it has temperature; δθ is the temperature of dummy variables; S2 is the heat
advantages of computational efficiency for large deformation and highly flux applied area; S3 is the applied convection area; [D] are the heat
non-linear problems [16,17]. During glass tempering process, not only conduction properties of the matrix material; and {v} is the conductivity
the dependence of temperature and stress on time is nonlinear, but of air.
also material properties change significantly as strain rate and temper- If the temperature change within the unit cell can be assumed to be
ature increase. The geometry and mathematical model [18] to be used represented by a polynomial, then the integral child domain decompo-
were established, and the boundary conditions for the simulations sition division unit can be expressed as follows:
were set on the basis of the actual toughening conditions.
T ¼ fNgT fθe g ð4Þ

2. Simulation analysis
where {N}T is the unit shape function and {θe} is the unit node temper-
ature vector. By considering the temperature in the integral equation
Glass tempering primarily involves blowing air on both sides of a
(Eq. (4)), obtain the following:
glass sample heated to the initial glass temperature T0 (usually approx-
imately 650 °C), in order to quench the glass sample to a certain temper- Z Z
ature (approximately 200 °C). Thus, mathematical models for the ρcfN gT fNgdðvolÞfθgþ ρc fNgT fvgT ðBÞdðvolÞfθg
vol Z vol Z Z
material and thermal boundary load are required to investigate the
þ fBgT ðDÞðBÞdðvolÞfθg ¼ fNgq dðS2 Þþ fNgλθB dðS3 Þ ð5Þ
glass tempering process. Calculations based on the material model Z vol S2 S3
parameters and the thermal boundary conditions can provide informa- − fNgT fN gλdðS3 Þ
tion on the history of the temperature and the stress changes occurring S3
during the glass toughening process, as well as on the magnitude and
distribution of the final residual stress. where [B] = {L}T ⋅ [N]. By taking into account the material properties,

Table 2
The thermal conductivity and the specific heat are characteristics varying with temperature as shown in this table [14].

Temperature (°C) 150 200 250 300 350 400 450 500 550 600 650
Specific heat Cρ(J(kg K)−1) 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437
Thermal conductivity λ(W(m K)−1) 1.14 1.19 1.24 1.29 1.34 1.39 1.44 1.49 1.54 1.59 1.64
74 B.-W. Fan et al. / Journal of Non-Crystalline Solids 437 (2016) 72–79

Fig. 1. Finite element mesh model and convective heat transfer model of the glass surface (uc: point of surface center, cc: point of model center).

Eq. (5) can be rewritten in the matrix form: where Sij is the deviating stress, eij is the deviating strain, Tref is the
n o reference temperature (here, T is the temperature, while τ denotes the
ðC ðθÞÞ θ_ þ ðK ðθÞÞfθg ¼ Q ðθ; tÞ: ð6Þ time), σkk is the positive stress along the x direction, εkk is the positive
strain along the x direction, G1 is the time-dependent shear modulus,
G2 is the time-dependent bulk modulus, and εth is the thermal strain.
Further, ξ (t) and ξ'(t) is reduced time, they can be expressed as follows:
2.2. Temperature field initial conditions and convection boundary
conditions Z t 0 Z τ 0
dt 0 dt
ξðt Þ ¼ 0 ; ξ ðt Þ ¼ 0 ð12Þ
First, the initial temperature of the cooling points is set. The relation- 0 Φ½T ðt Þ 0 Φ½T ðt Þ

ship between the coefficient for convective heat transfer of the medium
between the object and the surrounding environment and the temper- where Φ(t) is the shift function, which represents the temperature
ature of the medium can be expressed as follows: and time. In this simulation, the Arrhenius equation was used as the
temperature conversion equation; it can be expressed as follows [20]:
dθ  
−k ¼ λ θ−θ f ð7Þ   
dn 1 1
ΦðtÞ ¼ exp D − ð13Þ
T ref T
where n is the outer normal direction of the object boundary and λ is the
heat transfer coefficient, including for convective (λc) and radiation (λr)
where T denotes the instantaneous temperature and D is the material
heat transfer:
factor, which changes with a change in the reference temperature [21,
λ ¼ λc þ λr ð8Þ 22].
Viscoelastic glass can be represented using a generalized Maxwell
where in the average convective heat transfer coefficient associated stress relaxation model. In the generalized Maxwell model, the relaxa-
with the use of E.Pohlhausen's number of characteristics can be tion modulus of the material can be represented using a Prony series.
determined as follows [16]: Using the relaxation tensile and torsion modulus test data of the glass
for a constant reference temperature, the Prony series coefficients can
k k  1=3 be obtained [23].
λc ¼ Num ¼ 0:37 Re0:8
m −23500 Pr m ð9Þ
l l
2.4. Thermal boundary model
where Nu, l, Re, and Pr are the Nusselt number, glass sample length,
Reynolds number, and Prandtl number, respectively. Air with a temperature of 20 °C is blown on the two glass surfaces
The radiation heat transfer coefficient is expressed as follows using during the quenching procedure; this results in the convective transfer
the Stefan–Boltzmann law: of heat between the cold air and the surface, heat conduction within the
   glass, and heat radiation to the environment. The convective transfer of
λr ¼ σSB e θ2 þ θ2f θ−θ f ð10Þ heat is the main reason for the rapid cooling of the glass, and the convec-
tion heat transfer coefficient is assumed to be constant. Radiation
where σSB and e are the Stefan–Boltzmann constant and the blackness accounts for a small share of the overall calorific value, so the effect of
of the glass surface, respectively. radiation heat transfer can be ignored. In order to ensure that the simu-
lation conditions correspond to the actual working conditions, one
2.3. Mathematical model of material needs to consider the convective heat transfer coefficient for the border
around the glass sample. On the basis of the actual working conditions
Tempered glass exhibits temperature-dependent mechanical prop- as well as the relevant values given in the literature, the convective
erties similar to those of nonlinear elastic adhesives, and its stress and heat transfer coefficient of the two surfaces (h1) and four side bound-
strain change over time (relaxation and creep). Using the time–temper- aries (h2) were set. The thermoviscoelastic characteristics are given in
ature equivalent equation to describe the time variation in the relaxa- Table 1 [14,22]. The thermal conductivity and the specific heat are
tion modulus of the glass under different temperatures, then get the characteristics varying with temperature as shown in Table 2 [22].
following expression [19]:
Z 2.5. Finite element analysis model
t  0  deij ðτÞ
Sij ¼ G1 ξ−ξ ; T ref
Z0 dτ
ð11Þ The dimensions of the flat glass samples were 60 mm × 40 mm,
t   d
0 and their thicknesses were 1.8, 2, 3, 4, and 6 mm, respectively. A 1/8
σ kk ¼ G2 ξ−ξ ; T ref ½ε ðτ Þ−ε th ðτ Þdτ
0 dτ kk symmetrical geometric model of the specimens was used for the
 B.-W. Fan et al. / Journal of Non-Crystalline Solids 437 (2016) 72–79 75

analyses, as shown in Fig. 1. The grid unit used was a 60-node hexahe-
dron unit, since the border area of the glass samples and the stress
along their thickness direction changed significantly, in both parts
with a dense grid that more than 60 (as shown using circle “1” in
Fig. 1). On the surface of the model, the boundary conditions corre-
sponding to symmetric displacement were applied, as shown in Fig. 1.
First, an analysis of the quenching temperature field was performed.
Then, the temperature field boundary conditions for each time
increment were loaded as the boundary conditions for the stress field
analysis. To ensure that the analysis of the time step and the tempera-
ture field analysis for various time steps were accurate, subroutines
were used. Different total running times (300–600 s) were considered
for the glass samples with different thicknesses. The time-dependent
temperature conversion function was defined in a subroutine.
3. Results and discussion
3.1. Temperature field
Fig. 2 shows the temperatures at the centers and boundaries of the
glass samples as functions of the cooling time; these results are also
listed in Table 3. The red lines indicate the relationship between the
instantaneous temperature of the center point (Tuc) on the upper
surface and the quenching time, t, while the green lines represent the
change in the temperature of the center point (Tcc) on the central
plane. Further, the difference in the temperature, ΔT, (Tcc − Tuc) over
time is represented by the blue lines. The time at which the temperature
difference was the maximum (ΔTmax) is expressed as t(ΔTmax), while
the quenching period, Δt, is defined as the time over which the temper-
ature decreased from the initial value (650 °C) to 200 °C. It can be seen
from Fig. 2 and Table 3 that, with a decrease in the glass thickness,
ΔTmax, t(ΔTmax), and Δt reduced. ΔTmax has a small decrease from
76 to 65 °C when the thicknesses of glass samples decrease from
6 mm to 4 mm. However, it presents a sharp decrease when the glass
thickness is smaller than 3 mm. The values of ΔTmax are 31, 20 and
10 °C for 3, 2 and 1.8 mm-thick glass samples, respectively. Similarly,
it can be seen from the data in Table 3, the decrease in t(ΔTmax) is
not large (from 7.5 to 5.0 s) when the thicknesses of glass samples
decrease from 6 to 4 mm, in contrast, t(ΔTmax) decreases to 1.4, 0.15
and 0.1 s for 3, 2 and 1.8 mm-thick glass, respectively. Consequently,
for the 2 and 1.8-mm-thick glass samples, the quenching periods were
significantly lower. This would make it difficult to establish the temper-
ature gradient. Table 3 also lists quenching period Δt used in the actual
production for the 3–6 mm thick glass samples [24,25]. On comparing

Fig. 2. Temperatures at the surfaces and centers of the glass samples versus time (the red lines indicate the relationship between the instantaneous temperature of the center point (Tuc) on
the upper surface, the green lines represent the change in the temperature of the center point (Tcc) on the central plane, the blue lines represent the temperature difference (Tcc − Tuc)).
(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
76 B.-W. Fan et al. / Journal of Non-Crystalline Solids 437 (2016) 72–79

the values in Tables 3, it can be seen that the simulation results were in
good accordance with those obtained from the actual tempering
process. That is to say, the developed temperature field model was a
reasonable one. Further, it should also be applicable in the case of
glass samples 2 and 1.8 mm in thickness, which provides the tempera-
ture gradients for the study of stress gradient in the next step.
To further investigate the toughening of thin glass, we used different
surface heat transfer coefficients (160–200 W / (m2 K)) and different
broadside heat transfer coefficients (0–110 W / (m2 K)) in the next
simulation. It can be seen from Fig. 3 that for 4-mm thick glass sample,
the surface heat transfer coefficient, and, in particular, that for heat
transfer from the upper and lower surfaces has a significant effect on
the temperature gradient along the thickness direction. ΔTmax
increases from 40 to 65 °C when the surface heat transfer coefficients
increases from 160 to 200 W / m2 K. It can be seen that, when surface
heat transfer coefficient is small, not only the surface heat transfer, but
also the broadside heat transfer has an impact on the temperature
gradient. However, this was not the case for high surface heat transfer
coefficient condition. For 2-mm thick glass sample, when the surface
heat transfer coefficients increases from 160 to 200 W / m2 K, ΔTmax
just increases from 14 to 19 °C. If we increase airflow or decrease the
gas temperature, it can increase heat transfer. For example, when the
surface heat transfer coefficient is 400 W / (m2 K), the maximum
temperature gradients ΔTmax can reach 38 °C for 2 mm-thick glass, as
shown in Fig. 3a. But 400 W / (m2 K) is very large for actual production.
More importantly, although we can increase heat transfer to obtain
sufficient temperature gradient for thin glass, the quenching period is
still very short. Maybe it needs new cooling methods and equipment
to optimize the tempering process.
for different thick glass pieces are listed in Table 4. Seen from Table 4,
the compressive stress are about 130 MPa for 6, 4 and 3 mm-thick
glass samples, while it decreases to about 90 and 70 MPa for 2 and
1.8 mm-thick glass samples, which illustrates that the smaller the thick-
ness of the glass is, the lower the stress will be. Because t(ΔTmax) and Δt
decrease sharply when the glass thickness decrease from 4 mm to
2 mm, the compressive stress layer thickness which is directly affected
by the cooling rate consequently decrease greatly. Fig. 5(b) gives the
ration of compressive stress to compressive stress layer thickness,
which indicates a great enhancement when the glass is thinner than
4 mm. The large ratio of compressive stress to compressive stress
layer thickness will make the glass samples easy to be broken under
the external force.
The distribution curves for instantaneous stress in the thickness
direction of the glass samples reflect the changes in the stress in the
samples. Consider the 6-mm-thick glass sample as an example: from
1 s to 5 s, the stress in the glass sample relaxed to a significant degree,
owing to the fact that the temperatures of the surface and center were
in the transition temperature range. At 5 s and 10 s, the curve shows
that, from the sample surface to a depth of 1.5 mm, the temperature
dropped to a value lower than the transition temperature, and the
degree of stress relaxation was not substantial. Further, in this middle
part, the temperature was in the transition range. In addition, owing
to stress relaxation, the stress decreased gradually to zero. Then, 10 s
later, in this middle area, tensile stress was generated. Initially, the

3.2. Stress field

The highest stress relaxation point t (the same as the time at which
the temperature difference was the maximum t(ΔTmax)) is defined as
the time at which the stress relaxation rate of the glass is the largest.
The stress formation time Δt (the same as the quenching period) is
defined as the time needed for the stress to reach its final stable value.
Fig. 4(a) shows the dependence of the transition point t of the stress
growth on the glass sample thickness; it can be seen that the glass
sample thickness has a marked effect on the stress. This was also seen
in the case of the stress change time (Fig. 4(b)). When these results
are combined with those of the simulation of the temperature field,
for the same heat transfer coefficient (200 W / (m2 K)), it can be seen
that the time was sufficient for the stress in the glass samples with
thicknesses larger than 3 mm to reach the final stable value. However,
this was not the case for the samples with thicknesses lower than
3 mm. This is the reason it is difficult to temper thin glass samples.
From the viewpoint of a theoretical analysis, glass samples with a thick-
ness of 3 mm or less can be tempered by optimizing the tempering
process and equipment. Ensuring that the quenching time is accurate
is crucial.
Fig. 5(a) shows the final stress distributions for different thick glass
samples. Compressive stress and thickness of compressive stress layer

Table 3
Temperature field simulation results for glass with different thicknesses (h 1 =
200 W / (m2 K), h2 = 900 W / (m2 K)) and the quenching period Δt(s) used in the
actual production.

Thickness (mm) 6 4 3 2 1.8

Simulated ΔTmax 76 65 31 20 10
(°C)
Simulated t(ΔTmax)(s) 7.5 5.0 1.4 0.15 0.1
Simulated Δt(s) 50 25 6.5 4.6 2
Δt(s) used in the actual 50–60 15–20 5–10
Fig. 3. The influence of heat transfer coefficient on the ΔTmax values of the (a) 2-mm-thick
production [24,25] [24,25] [24,25]
and (b) 4-mm-thick glass samples.
 B.-W. Fan et al. / Journal of Non-Crystalline Solids 437 (2016) 72–79
Fig. 4. (a) Highest stress relaxation point and (b) stress formation time.
temperature of this glass sample in the middle region was still in the
transition temperature range. Thus, there was certain degree of tensile
stress relaxation. As a result, the tensile stress distribution curve at 5 s
and 10 s was concave. Between 10 s and 50 s, the glass temperature
was out of the transformation temperature range, and the degree of
stress relaxation could be ignored. Thus, the transient stress is produced
by the strain which is generated by structural relaxation and thermo-
elastic strain. It was found that this phenomenon significantly reduces
the temperature difference 10 s after glass tempering. Therefore, in
the middle area, the rates of increase of the tensile stress and the surface
compression stress decreased gradually till the stresses plateaued. The
surface area decreased, forcing the middle region to contract. In this
case, the temperature of the intermediate zone was still high, and the
sample exhibited a certain degree of viscosity, as the stress was relaxed
to a small degree. Further, as the temperature dropped, the surface
began to solidify. Shrinkage was observed in the middle area with the
formation of constraints, which eventually produced an intermediate
region with a tensile stress and a surface compressive stress. Fig. 6
also shows the changes in the stress for the 1.8 and 2-mm-thick molded
glass samples. Thermoviscoelastic deformation in a very short period
caused the structural relaxation to occur to only a small degree; this
was an important factor in the generation of the instantaneous stress.
Further, the fact that the samples underwent consolidation on the sur-
face and internally at the same time made it difficult to fully temper
and heat-strengthen them.
77
Fig. 5. (a) Stress gradients for the glass samples with different thicknesses at the center
point and (b) stress-to-thickness ratio.
4. Summary
Using the viscoelastic stress relaxation theory and the theory of
relaxation of the glass structure proposed by Narayanaswamy, the
actual glass and tempering thermal boundary conditions were set for
simulating the physical quenching of glass samples with different thick-
nesses. With a decrease in the glass thickness, the difference in the tem-
perature of upper and center point ΔTmax, the time at which the
temperature difference was the maximum (ΔTmax) t(ΔTmax), and
quenching period Δt reduced, especially they present a sharp decrease
when the glass thickness is smaller than 3 mm. This would make it
difficult to establish the temperature gradient in thin glass. Different
from thick glass, not only surface but also broadside heat transfer has
obvious influence on the temperature gradient for glass thinner than
2 mm. For the same heat transfer coefficient (200 W / (m2 K)), it can
be seen that the time was sufficient for the stress in the glass samples
with thicknesses larger than 3 mm to reach the final stable value. This
was not the case for the samples with thicknesses lower than 3 mm.
Although we can increase heat transfer to obtain sufficient temperature
Table 4
Compressive stress and thickness of compressive stress layer.
Thickness (mm) 6 4 3 2 1.8
Compressive stress (MPa) 135 126 120 90 70
Thickness of compressive stress layer (mm) 1.4 0.8 0.5 0.35 0.28
78 B.-W. Fan et al. / Journal of Non-Crystalline Solids 437 (2016) 72–79
Fig. 6. Instantaneous stresses in the glass samples with different thicknesses.
gradient for thin glass, the quenching period is very short. Further, the
fact that the samples underwent consolidation on the surface and inter-
nally at the same time made it difficult to fully temper and heat-
strengthen them. However, glass samples with a thickness of 3 mm or
less can be tempered by optimizing the tempering process and
equipment.
Acknowledgments
This work was supported by the National Natural Science Founda-
tion of China (51335002, 51272033), the Priority Academic Program
Development of Jiangsu Higher Education Institutions.
References
[1] X.H. Zhang, H. Hao, Z.Q. Wang, Experimental investigation of monolithic tempered
glass fragment characteristics subjected to blast loads, Eng. Struct. 75 (2014)
259–275.
[2] G.Z. Shao, Q.S. Wang, H. Zhao, Maximum temperature to withstand water film for
tempered glass exposed to fire, Constr. Build. Mater. 57 (2014) 15–23.
[3] A. Koikea, S. Akibaa, T. Sakagamia, K. Hayashia, Difference of cracking behavior due
to Vickers indentation between physically and chemically tempered glasses, J. Non-
Cryst. Solids 24 (2012) 3438–3444.
[4] H. Loch, D. Krause, Mathematical Simulation in Glass Technology, Springer, Berlin,
2002.
[5] T.F. Soules, R.F. Busbey, S.M. Rekhson, A. Markovsky, M.A. Burke, Finite element
calculation of stresses in glass parts undergoing viscous relaxation, J. Am. Ceram.
Soc. 70–2 (1987) 90–95.
[6] H. Aben, C. Guillermet, Photoelasticity of Glass, Springer-Verlag, Berlin, 1993.
[7] A. Kuske, G. Robertson, Photoelastic Stress Analysis, Wiley, London, 1974.
[8] C.L. Shepard, B.D. Cannon, M.A. Khaleel, Measurement of internal stress in glass
articles, J. Am. Ceram. Soc. 86–8 (2003) 1353–1359.
[9] G.W. Brodland, A.T. Dolovich, Curved-ray technique to measure the stress profile in
tempered glass, Opt. Eng. Bellingham. 39–9 (2000) 2501–2505.
[10] G.T. Moynihan, A.J. Easteal, M.A. DeBolt, Dependence of the fictive temperature of
glass on cooling rate, J. Am. Ceram. Soc. 59 (1–2) (1976) 12–15.
[11] O.S. Narayanaswamy, A model of structural relaxation in glass, J. Am. Ceram. Soc.
54–10 (1971) 491–498.
[12] A. Tool, Relation between inelastic deformability and thermal expansion of glass in
its annealing range, J. Am. Ceram. Soc. 29 (9) (1946) 240–253.
[13] O.S. Narayanaswamy, Stress and structural relaxation in tempering glass, J. Am.
Ceram. Soc. 61 (1978) 146–152.
[14] J.H. Nielsen, J.F. Olesen, P.N. Poulsen, Finite element implementation of a glass
tempering model in three dimensions, Comput. Struct. 88 (2010) 963–972.
[15] ABAQUS User Manual, Version 6.3, Hibbitt, Karlsson & Sorensen, Inc., 2002
[16] P.J. Arrazola, T. Özel, Investigations on the effects of friction modeling infinite
element simulation of machining, Int. J. Mech. Sci. 52 (2010) 31–42.
 B.-W. Fan et al. / Journal of Non-Crystalline Solids 437 (2016) 72–79
[17] D. Deng, H. Murakawa, Numerical simulation of temperature field and residual
stress in multi-pass welds in stainless steel pipe and comparison with experimental
measurements, J. Comp. Mater. Sci. 37 (2006) 269–277.
[18] M.B. Boubakera, B.L. Correb, Y. Meshakaa, Finite element simulation of the slumping
process of a glass plate using 3D generalized viscoelastic Maxwell model, J. Non-
Cryst. Solids 405 (2014) 45–54.
[19] R. Gordon, Thermal tempering of glass, J. Elast. Strength Glas. 5 (2012) 167–186.
[20] G. Robert, O.S. Narayanaswamy, Stress and volume relaxation in annealing flat glass,
J. Am. Ceram. Soc. 53 (7) (1970) 380–385.
[21] E.H. Lee, T.G. Rogers, T.C. Woo, Residual stresses in a glass plate cooled symmetrical-
ly from both surfaces, J. Am. Ceram. Soc. 48 (9) (2006) 480–487.
79
[22] L. Daudeville, H. Carre, Thermal tempering simulation of glass plates: inner and edge
residual stresses, J. Therm. Sresses 21 (6) (1998) 667–689.
[23] T.F. Soules, R.F. Busbey, S.M. Rekhson, Finite element calculation of stresses in glass
parts undergoing viscous relaxation, J. Am. Ceram. Soc. 70 (2) (1987) 90–95.
[24] W.G. Xu, Proces parameters setting of strengthened glass, J. Glass Enamel 39 (4)
(2011) 27–28.
[25] W.G. Xu, Control process technical parameters of strengthened glass, J. Archit. Funct.
Glass 11 (2011) 24–26.