Documente Academic
Documente Profesional
Documente Cultură
INTRODUCTION
Segmental construction is an appealing choice to many
bridge designers and owners. It makes achieving longer span
lengths possible while still handling and transporting
manageable size segments (Fig. 1). The prestressed (PS)
concrete box girder is one of the segmental alternatives that
has been gaining popularity in the past 30 years as it becomes
more competitive to other alternatives. Box sections provide
favorable structural properties such as high flexural stiffness
and superior torsional rigidity. They can also be used to
construct variable depth spans that are aesthetically Fig. 2—Stress block from web of PS segmental bridge.
appealing. Cracking of PS concrete segmental box girder
bridges negatively impacts their durability; therefore, they
are designed to be free of cracks under service conditions. direction σx (refer to Fig. 1 for coordinate system). They are
Designers are often faced with the challenge of estimating an caused by straining actions due to gravity loads, long-term
allowable tensile stress that can achieve this goal. Allowable effects, and post-tensioning (PT) forces. In regions with high
tensile stresses for consideration at the top and bottom fibers shear demands, another normal stress σy may exist if vertical
of the girder are well established in design codes.1-3 They PT bars are used in the webs. Shear stresses v are caused by
vary based on several factors such as the type of joint direct shear forces as well as torsional effects. This state of
(Type A or B). These stresses are usually provided as a stress (Fig. 2) is somewhat different than the state of stress at
function of the compressive strength fc′ and normally range the top and bottom fibers in that the critical stress that may
from 6 f c′ to zero tension. In some cases, such as the case cause cracking is the principal stress σ1, as can be demonstrated
for Type B joints with external tendons, AASHTO- by Mohr’s circle (Fig. 3). Controlling shear cracking
Segmental2 calls for a minimum compressive stress of 100 psi. requires that the principal stress σ1 be limited to an allowable
By complying with the aforementioned allowable tensile tensile stress ft,all. AASHTO-Segmental2 does not offer
stresses, the possibility of flexural cracking at the top and guidance to designers with regard to ft,all in webs, which
bottom fibers should be eliminated. These allowable stresses
are only good for the extreme fibers and are not applicable to ACI Structural Journal, V. 103, No. 4, July-August 2006.
MS No. 04-197 received June 2, 2005, and reviewed under Institute publication policies.
webs. In webs, cracks may form due to a biaxial state of Copyright © 2006, American Concrete Institute. All rights reserved, including the making
stress resulting from a combination of shear and normal of copies unless permission is obtained from the copyright proprietors. Pertinent discussion
including author’s closure, if any, will be published in the May-June 2007 ACI
stresses. The normal stresses are usually in the longitudinal Structural Journal if the discussion is received by January 1, 2007.
σ tu σ cu
Each of the expressions is given in MPa and psi. The plots ------- = 1 + 0.85 -------
- (4)
show that the difference between these expressions is not f tu f c′
large except at the ends of the plotted range, that is, low and
high compressive strengths. It is clear, however, that the Equation (4) is a slightly modified version of the one
third expression (Eq. (3)) provides a better estimate of the proposed by Kupfer and Gerstle,7 where the second term
tensile strength over a wider range of concrete compressive coefficient is equal to 0.8. This modification reflects a better
Resistance model
A generic resistance model that is good for different
compressive strengths fc′ and level of corresponding
compressive stress σcu/fc′ can now be written by combining
Eq. (3) and (4)
σ cu
× 1 + 0.85 -------
0.69
1.38 ( f c′ ) - , in psi
f c′
σ tu = (5)
σ cu
× 1 + 0.85 -------
0.69
0.30 ( f c′ ) - , in MPa
f c′
Equation (3) is chosen as the basis of the general tensile Fig. 7—Cross section of PS bridge used for calibration study.
strength expression for two reasons. First, it is more appropriate
to compare the tensile principal stress at the centroidal axis
of a PS segmental box to the splitting tensile strength than the along this interchange are used to study the state of stress in
modulus of rupture. Shear stresses are higher at the centroidal webs of PS segmental bridges. They also form the basis for
axis than they are at the extreme fibers, which translates into developing a rational, allowable tensile stress for use in
higher principal stresses σ1. Hence there is no clear stress controlling web cracking. The proposed bridge designs are
gradient, as is the case in regions controlled by flexure. considered typical segmental box girder bridge in today’s
Second, it is clear that Eq. (3) provides better estimates over practice. Figure 7 shows the cross sections of the bridges,
a wide range of concrete compressive strengths (Fig. 4). which are constant in depth (2.60 m [8 ft 6.5 in]). Additional
Therefore, Eq. (5) is used in the calibration of the allowable information about some of the bridges’ major geometric
tensile stress taking into account the uncertainties involved parameters is listed in Table 2. The interchange included
as will be seen later. It should be noted that Eq. (5) was three two-span bridges, one three-span bridge, and two 10-span
developed for the range of compressive strengths covered in bridges. All bridges were horizontally curved with a radius
this study and may not be applicable to high-strength or of curvature equal to 236.22 m [775 ft]. The bridges were
high-performance concrete without further investigations. designed to be constructed using the balanced cantilever
method and are post-tensioned using internal tendons. The
STATE OF STRESS IN WEBS main materials specified for the bridges are: concrete
OF PS SEGMENTAL BRIDGES strength fc′ = 6000 psi, and low relaxation strands fpu = 270 ksi.
A highway interchange in Florida is planned to be The analyses were performed using BRUCO14 taking
constructed in the near future. Several design alternatives into account the construction steps as specified in the project
were considered, including PS concrete segmental box plans. HL-93 loads,3 which is the combined effect of lane
girder bridges. Stress data from the design of six bridges loading and the maximum of the effects of a standard truck
S5 0.81 1.69 –8.6 –1.6 2.09 0.23 CALIBRATION OF ALLOWABLE TENSILE STRESS
(118) (246) (–1260) (–246) Allowable tensile stress
S6 1.22 1.07 –5.9 –1.0 0.88 0.15 As stated previously, designers usually compare the principal
(178) (156) (–861) (–155) tensile stress σ1 to an expression of the form, constant × f c′ ,
S7 0.92 2.66 –3.3 –2.6 2.87 0.13 where the constant is between 0.25 and 0.33 MPa (3.0 and
(135) (387) (–484) (–386)
4.0 psi). This form of ft,all will be one of three expressions
calibrated in this study. The other expressions will be
Table 4—Statistical properties of variables attempts to enhance the ability to estimate an allowable
involved in study tensile stress over a wide range of design variables. The three
Current study
expressions are
Coefficient of
Variable Bias variation 0.5
f t, all = κ 1 ( f c′ ) (6)
Dimensions (h, d, b) 1.00 3.0%
Area of strands (Aps) 1.00 1.5%
0.5
Concrete Strength ( f ′c ) 1.10 18.0% f t, all = λ σ2 × κ 2 ( f c′ ) (7)
Uncertainty in concrete tensile strength (ftu) 1.00 *
0.44 MPa
0.7
Uncertainty in biaxial failure envelope ( λ σ2 ) 1.00 10.6% f t, all = λ σ2 × κ 3 ( f c′ ) (8)
Model uncertainty (α) 1.01 4.5%
Component dead load (DC) 1.05 10.0% where λσ2 is a dimensionless modifier to account for the
Wearing surface dead load (DW) 1.10 20.0% corresponding compressive stress σ2 based on the regression
HL-93 1.25 18.0% study presented previously (Eq. (4)).
Live load Impact (IM) 0.10 18.0%
Traffic† σ
λ σ2 = 1 + 0.85 -----2
varies —
(9)
*
Value shown is standard deviation. f c′
†
1.05 for two-lane bridges (S1, S2, S4, S5, and S6) and 0.90 for three-lane bridge
(S7) assuming ADTT > 5000.16
The first expression (Eq. (6)) is currently used by most
designers, and the aim is to evaluate a calibrated constant κ1
and a load tandem, were considered. The number of design that results in a probability of failure Pf in line with values
lanes was determined as per AASHTO-LRFD.3 Accordingly, targeted by the design code. In Eq. (7), the effect of the prin-
each bridge was loaded by two lanes filled with HL-93 cipal compressive stress σ2 on the tensile strength of concrete
loading except for Bridge S7, which was loaded with three is introduced to Eq. (6). The last expression (Eq. (8)) is a
lanes. Therefore, Bridge S7 live load envelopes were modified diversion from the historical fc′ power exponent of 0.5, as it
using a multiple presence factor of 0.85. The loads were explores a power exponent equal to 0.7, which has been
positioned such that the maximum torsional effects are already adopted by other design codes.12 This is an attempt
produced. The torsional effects add to the shear stresses due to improve the ability to estimate a more reliable tensile
to flexural shear forces. The truck portion of live load effects stress over a wider range of concrete strengths (Fig. 4).
was increased by 33% to account for the dynamic allowance
(IM) according to AASHTO-LRFD. Finally, Service III
Limit state function for calibration of ft,all
limit state was considered, which assigns a load factor A limit state (performance) function Z, based on the
γLL+IM equal to 0.8 for live load effects. The remaining load demand and capacity, is first set for the calibration study
factors were taken as unity. The principal stresses in each
analysis segment (element) were computed and the most
critical segment in each bridge was identified. It should be Z = σ tu – α ( σ 1, DC + σ 1, DW + σ 1, PT + σ 1, LL ) (10)
noted that pier segments were not considered when identifying
the most critical sections because they are provided with a where σtu is the tensile strength of concrete, and σ1,DC,
rigid diaphragm. σ1,DW, σ1,PT, and σ1,LL are the principal tensile stresses due
In preparation for the calibration study, the effect of each to component dead loads, wearing surface dead load, post-
stress component was isolated. For example, Fig. 3 shows tensioning, and live load, respectively. Each of these variables
how the portion of the tensile principal stress caused by live is treated as a random variable following the statistical
loads was isolated from the dead load and prestressing properties (bias and coefficient of variation) given in Table 4.
effects. Stresses caused by prestressing effects and dead The information listed in Table 4 has been obtained from
Pf = Φ(–β) (11)
Parametric study
Due to the limited size of actual bridge designs available
for the study, a virtual design space that covers a wide range
of variables influencing the tensile capacity of concrete in
segmental bridges will now be studied. Table 3 shows that
the ratio between the live load plus impact (LL + IM) and
dead load (DL) components of the principal stress σ1 varies
between 0.52 and 2.87. A σ1,LL+IM/σ1,DL ratio between 0.0
and 4.0 is investigated. The compressive principal stress was
also investigated by varying the ratio σ2/fc′ between 0.0 and
0.30. Concrete strengths fc′ between 35 and 55 MPa (5 and
Fig. 10—Comparison between proposed expressions for
8 ksi) are also investigated.
κ1 = 0.25 MPa (3.0 psi),κ2 = 0.29 MPa (3.5 psi), κ3 =
Figure 10 shows the reliability index resulting from each
0.13 MPa (0.6 psi), and fc′ = 42 MPa (6000 psi).
of the proposed expressions over a range of σ2/fc′ ratio. The
three lines plotted for each expression correspond to
different (LL + IM)/DL ratios, namely, 0.23, 0.58, and 1.67.
As expected, higher live load ratios result in lower reliability
because of the higher uncertainty inherent in live loads (refer
to Table 4). The reliability index values plotted in this figure,
however, are still within acceptable range and can be
increased or reduced by slightly changing the constants κ1,
κ2, and κ3. The more important observation is the better
uniformity of results obtained using Eq. (7) and (8). Over the
shown range of σ2/fc′ , the reliability index drops an average
of 1.7, which translates to an order of magnitude when
converted into a probability of failure Pf . This implies that
Eq. (6) will deliver overconservative designs for situations
with low σ2/fc′ ratios and unconservative designs for high
σ2/fc′ ratios. The introduction of the modifier λσ2 in Eq. (7)
Fig. 11—Comparison between proposed expressions for and (8) reduces this inconsistency and produces reliability
κ1 = 0.25 MPa (3.0 psi),κ2 = 0.29 MPa (3.5 psi), κ3 = 0.13 MPa index values that drop an average of 0.48 and 0.50, respectively.
(0.6 psi), and σ2 / fc′ = 0.15. The effect of concrete compressive strength is illustrated
in Fig. 11. Equation (8) demonstrates a more uniform design
is mainly due to the fact that it is the only bridge where the over the range of fc′ plotted in the figure. The other expressions,
critical section falls in a region provided with additional web Eq. (6) and (7), show a much larger difference between the
PT bars. This additional prestressing force increases the reliability index for designs conducted using fc′ = 35 and 55 MPa
compressive principal stress σ2 and, hence, reduces the (5 and 8 ksi). This can be attributed to the different power
tensile strength more than other bridges. This observation is exponent (0.7) used in Eq. (8), which matches experimental data
not as severe for the calibration of κ2 (Fig. 9(b)) and even better than the traditional power exponent of 0.5.
less so for the calibration of κ3 (Fig. 9(c)), which is expected,
as both expressions account for the effect of σ2 through the SUMMARY AND CONCLUSIONS
modifier λσ2. The allowable tensile stress of concrete ft,all is calibrated
for use in web design of PS segmental box girder bridges.
Choice of κi The proposed limit is intended for controlling cracking in
The plots in Fig. 9 clearly demonstrate that choosing κ-values nondisturbed regions, that is, B-regions. The first-order
that would result in a constant target reliability index across reliability method (FORM) is used to conduct the calibration.
different scenarios is impossible. Therefore, the choice has The first (Eq. (6)) of three calibrated expressions is currently
to be one that produces values within a range acceptable to used by most designers and ignores the adverse effect of the
code committees. This acceptable range depends on many compressive principal stress on concrete tensile strength.
factors, such as: consequences of failure, importance of Two expressions (Eq. (7) and (8)) that account for this effect
component, mode of failure, and purpose of analysis (design are also calibrated. One of these expressions (Eq. (8)) uses a
versus load rating). The data presented in Table 5 and Fig. 9 power exponent for fc′ of 0.7, which has proven more accurate
provide information that will assist code committees in in light of experimental data covering a wide range of
choosing an expression that matches their design philosophy concrete compressive strengths. The calibration is performed
and target reliability index. for a range of reliability index β, 2.50 to 3.50, which is
To demonstrate the differences between the proposed considered acceptable for service limit states.
expressions, values of 0.25, 0.29, and 0.13 MPa (3.0, 3.5, A parametric study is conducted to illustrate the performance
and 0.6 psi) are assumed for κ1, κ2, and κ3, respectively. of the calibrated expressions. The parametric study is for a
Figure 9 shows that the proposed values (horizontal lines) chosen set of constants, κ1, κ2, and κ3, that would result in a
would result in reliability indexes around 3.0 for most of the reliability index in line with current code targets. Code
analyzed bridges. The following study is conducted to committees may choose higher or lower targets by modifying