Reliability-based sensitivity

analysis for prestressed

concrete girder bridges

Anna M. Rakoczy and Andrzej S. Nowak

he new generation of load and resistance factor

design (LRFD) codes is based on probabilistic
methods. Structural performance depends on the
uncertainty involved in the applied load and in the loadcarrying capacity. Therefore, the load and resistance
factors represent partial safety factors and their values are
a function of bias factors (ratios of mean to nominal) and
coefficients of variation.1,2

Over the past 30 years, both material properties and bridge

loads have changed.
The objective of this paper is to present the results of a reliability analysis for prestressed concrete girders using the most
recent live load and resistance models. Girder spans ranging
from 80 to 200ft (24 to 61m) and girder spacings of 8 and
10ft (2.4 and 3m) are included.
The paper also presents a derivation of sensitivity functions
for various parameters that affect performance of prestressed
concrete girders. The developed procedure is demonstrated on
representative girders.

The primary load combination for highway bridges

includes dead load, live load, and dynamic load. Each of
the individual components is random in nature. Moreover,
the live-load effect also depends on truck position on the
bridge and load distribution factor. The statistical parameters of live load are taken from the recent study presented
by Rakoczy3 and Nowak et al.4 based on an extensive
weigh-in-motion survey.
The uncertainty in the resistance model is caused by variability in material properties, dimensions, and analytical
procedures. Each parameter can be considered as a random
variable and can be described by statistical parameters
such as the mean value , standard deviation , coefficient
of variation V, probability density function, and cumulative distribution function (CDF). Assessing the statistical
parameters requires extensive material and component
testing. However, because of the prohibitive cost, full-scale

PCI Journal | Fa l l 2013

81

-2
-3

2
1

400

350

300

250

200

-1

fpu , ksi
150

50

400

350

300

250

200

150

100

-1

fpu , ksi
50

Total (33,387)

Standard normal variable

Standard normal variable

Strands 0.5 in.

Source 1 (3908)
Source 2 (1158)
Source 3 (268)
Source 4 (9795)
Source 5 (18,258)

100

Strands 0.5 in.

-2
-3
-4

-4

-5

Data plotted separately for each source

All data plotted together

Figure 1. Cumulative density functions of tensile strength of 0.5in. diameter prestressing strands. Source: Reproduced from Nowak and Rakoczy (2012). Note: fu =
ultimate strength of reinforcing steel. 1in. = 25.4mm; 1ksi = 6.895MPa.

tests are limited or impossible. Therefore, the behavior of

large bridge components must be evaluated using analytical methods and simulations.
The objective of this study is to calculate the reliability indices for prestressed concrete girders using new statistical
parameters for material and load components. The sensitivity analysis was performed and the results presented for
resistance and load parameters that can affect reliability
indices of prestressed concrete beams. The analysis is
presented for American Association of State Highway and
Transportation (AASHTO) type girders5 and Nebraska
University (NU) girders.6,7

Research significance
Prestressed, precast concrete beams are widely used in
highway multigirder bridges because of lower construction
costs and increased span length. Recently, new and updated
statistical parameters are available for material properties
and live load on highway bridges. Therefore, there is a
need to evaluate the performance of prestressed concrete
beams based on new load and resistance models. The
sensitivity analysis performed establishes a relationship between resistance and load parameters and reliability index
and thus helps to identify the most critical parameters for
prestressed concrete beams. The results of this study can
serve as a basis for verification of the current design code
provisions, including review of load and resistance factors.

Statistical parameters
of material
The statistical parameters of the material factor for tensile
strength of prestressing strands were developed based on
new data that include 47,421samples of strands of 0.5in.

82

Fall 2 0 1 3 | PCI Journal

(13mm) and 0.6in. (15mm) diameter. The data were

obtained from five U.S. fabricators.
For a more efficient interpretation, the statistical data were
plotted on normal probability paper. The properties and use
of normal probability paper can be found in textbooks on
statistics and probability theory, for example, Nowak and
Collins.8 The most valuable characteristic of probability
paper is easy evaluation of the statistical parameters as
well as type of the distribution function. The vertical axis
on normal probability paper is the inverse of the standard
cumulative distribution and represents the number of
standard deviations from the mean value. If the CDF curve
is close to a straight line, then the considered variable has
a normal distribution. In this case, the mean value and
standard deviation can be read directly from the CDF plot.
The mean value is the horizontal coordinate of intersection
with the CDF and the slope of the CDF is the inverse of
standard deviation. Bias factor is defined as the ratio of
the mean to nominal value, and the coefficient of variation
is calculated as a standard deviation divided by the mean.1,2
First, the CDF of the ultimate strength of prestressing
strands is plotted on normal probability paper. The CDFs
of tensile strength of prestressing strands of 0.5 and 0.6in.
(13 and 15mm) diameters are close to a straight line
(Fig.1 and 2). This is more visible when all of the data
points are plotted together.
The statistical parameters such as mean value , bias factor
, and coefficient of variation V can be determined directly
from the graph representing the CDF. Table1 summarizes
the statistical parameters for prestressing strands.
The statistical parameters for ordinary concrete were
obtained by Nowak et al.,9 including the test data of 28-day

-2

2
1
400

350

300

250

-1

200

fpu , ksi
100

50

400

350

300

250

200

150

100

-1

50

fpu, ksi

Total (14,028)

Standard normal variable

Standard normal variable

Strands 0.6 in.

Source 1 (700)
Source 2 (785)
Source 3 (212)
Source 4 (3442)
Source 5 (8895)

150

Strands 0.6 in.

-2

-3

-3

-4

-4

Data plotted separately for each source

All data plotted together

Figure 2. Cumulative density functions of tensile strength of 0.6in. diameter prestressing strands. Source: Reproduced by permission from Nowak and Rakoczy
(2012). Note: fu = ultimate strength of reinforcing steel. 1in. = 25.4mm; 1ksi = 6.895MPa.

compressive strength of concrete for standard cylinders,

6in. 12in. (150mm 300mm), ordinary concrete with
nominal compressive strength
of 3000 to 6500psi (21
to 45MPa), and high-strength concrete with nominal compressive strength
of 7000 to 12,000psi (49 to 84MPa).

The CDFs for each nominal compressive strength

from
3000 to 12,000psi (21 to 84MPa) were plotted on normal
probability paper (Fig.3). Table2 lists the recommended
bias factor and coefficients of variation V of compressive
strength . The statistical parameters are given separately

Table 1. Summary of the statistical parameters for prestressing

Strand diameter, in.

Number of samples

Bias factor

Coefficient of variation V

0.5

33,387

1.04

0.017

0.6

14,028

1.02

0.015

Source: Reproduced from Nowak and Rakoczy (2012). Note: 1in. = 25.4mm.

2
1

-2
-3
-4

fc = 3000 psi
fc = 4000 psi
fc = 5000 psi
fc = 6000 psi
fc = 7000 psi
fc = 9000 psi
fc = 12,000 psi

20,000

18,000

16,000

14,000

12,000

10,000

8,000

6,000

4,000

-1

2,000

f'c , psi
0

Standard normal variable

fc = 3500 psi
fc = 4500 psi
fc = 5500 psi
fc = 6500 psi
fc = 8000 psi
fc = 10,000 psi

Figure 3. Cumulative density functions of compressive strength of concrete. Source: Reproduced by permission from Nowak et al. (2011). Note:
compressive strength of concrete at 28days. 1psi = 6.895kPa.
PCI Journal | Fa l l 2013

= specified

83

Table 2. Recommended statistical parameters for compressive strength of concrete

Compressive strength

Compressive strength
of concrete f c' , psi

Shear strength

Bias factor

Coefficient of variation V

Bias factor

Coefficient of variation V

4000

1.24

0.150

1.24

0.180

5000

1.19

0.135

1.19

0.160

6000

1.15

0.125

1.15

0.150

7000

1.13

0.115

1.13

0.140

8000

1.11

0.110

1.11

0.135

9000

1.10

0.110

1.10

0.135

10,000

1.09

0.110

1.09

0.135

12,000

1.08

0.110

1.08

0.135

Note: 1psi = 6.895kPa.

for compressive strength and shear strength. It was assumed that the bias factor for shear strength is the same
as for compressive strength. The coefficient of variation V
of shear strength was assumed to be larger than that for the
corresponding
by 20%.10

coefficient of variation V of fy is 0.03.9

Resistance models
The resistance of a structural component R is a random
variable related to uncertainties resulting from material
properties and dimensions. It is convenient to consider R as
a product of nominal value Rn, material factor M, fabrication factor F, and professional factor P as is expressed in
Eq.(1).

Nowak et al9 analyzed the test data for steel reinforcing

bars with a yield strength fy of 60ksi (420MPa) and a wide
range of bar sizes from no.3 through no.14 (10M through
43M). The CDFs of the yield strength of reinforcing steel
fy for various bar sizes were plotted on normal probability
paper (Fig.4). The bias factor for reinforcing steel bars
is in the range of 1.12 to 1.14 for most diameters, except
no.3 reinforcing bars, for which the bias factor is 1.18.
The coefficient of variation of fy is small, between 0.02 and
0.04. The recommended bias factor for fy is 1.13, and the

R = RnMFP (1)

The nominal capacity of prestressed concrete girders was

determined according to AASHTO LRFD Bridge Design
Specification provisions for flexure. The formula for bend-

2
1

110

100

90

80

70

60

50

40

30

20

fu , ksi
10

Standard normal variable

-1
-2
-3
-4

No. 3
No. 5
No. 7
No. 9
No. 11

No. 4
No. 6
No. 8
No. 10

Figure 4. Cumulative density functions of yield strength of reinforcing bars. Source: Reproduced by permission from Nowak et al. (2011). Note: fu = ultimate strength
of reinforcing steel. No.3 = 10M; no.4 = 13M; no.5 = 16M; no.6 = 19M; no.7 = 22M; no.8 = 25M; no.9 = 29M; no.10 = 32M; no.11 = 36M; 1ksi = 6.895MPa.
84

Fall 2 0 1 3 | PCI Journal

ing moment Eq.(2) is based on AASHTO LRFD specifications (Eq.[5.7.3.2.2-1])5 provisions.

(2)

where
Aps = area of prestressing steel
fps = average stress in prestressing steel at nominal bending resistance
dp = distance from extreme compression fiber to centroid
of prestressing tendons
a = depth of the equivalent stress block = c1
c = distance from extreme compression fiber to neutral
axis

1 = stress block factor

As = area of mild steel tensile reinforcement
fs = stress in mild steel tensile reinforcement at nominal
flexural resistance
ds = distance from extreme compression fiber to centroid
of mild steel tensile reinforcement
= area of mild steel compressive reinforcement
= stress in mild steel compressive reinforcement at
nominal flexural resistance
= distance from extreme compressive fiber to the centroid of mild steel compressive reinforcement
b = width of compression face of member; for a flange
section in compression, effective width of flange
bw = web width or diameter of a circular section
hf = compression flange depth of an I or T member
The new material test data, described in the previous
section, served as a basis for material factor M. It was
observed that statistical parameters for material improved
over the years.
Variation in dimensions and geometry is represented by
fabrication factor F. No new data are available for the
fabrication factor; therefore, the recommended statistical

parameters were based on previous studies by Ellingwood

et al.11 For the dimensions of cast-in-place concrete components, the recommended parameters of the F bias factor
F and F coefficient of variation VF are 0.92 and 0.12,
respectively, for effective depth of slab and F of 1.01 and
VF of 0.04 for effective width of beam. For effective depth
of prestressed concrete girder, F of 1.0 and VF of 0.025 are
recommended.
For steel components, reinforcing bars, and stirrups, the
statistical parameters are F of 1.0 and VF of 0.01. The area
of reinforcing steel As is treated as a practically deterministic value, with F of 1.0 and VF of 0.015; similarly, the area
of prestressing strands has a F of 1.0 and VF of 0.01.
The professional (analysis) factor P represents the variation in the ratio between the actual resistance and what can
be analytically predicted using accurate material strength
and dimension values. The professional factors bias factor
P and coefficient of variation VP for a prestressed concrete
beam in flexure are 1.01 and 0.06, respectively. Those
statistical parameters of P are based on the study by Ellingwood et al.11
In this study, the statistical parameters of resistance were
taken from previous research.12 Nominal capacity was
calculated according to AASHTO LRFD specifications.
Then, each of the design parameters, which were treated as
random variables, was simulated 1million times using the
Monte Carlo technique, and then values of resistance were
calculated accordingly. The resulting CDFs were plotted on normal probability paper. The mean resistance R,
standard deviation of resistance R, bias factor of resistance R, and coefficient of variation of resistance VR were
calculated using CDF plots. The analysis was performed
for the NU girders and AASHTO girders with different
span lengths, different spacing between the beams, and
two types of concrete:
of 10,000psi (69MPa) for
prestressed concrete girders and
of 5000psi (34MPa)
for the concrete slab. Statistical parameters of resistance
were about the same for all types of girders. Statistical
parameters of flexural resistance for prestressed concrete
beams depended only on the strand diameter. The bias factor for resistance was 1.05 for strands of 0.6in. (13mm)
diameter and 1.07 for strands of 0.5in. (15mm) diameter.
The coefficient of variation for both cases was low, about
0.07. The bias factor for shear resistance for prestressed
concrete beams was 1.20 and coefficient of variation was
0.115 regardless of the size of the strands. More detailed
information about the statistical parameters is provided by
Nowak and Rakoczy.12

Load and load combination

models
The main load components for short- and medium-span
highway bridges are dead load, live load, and dynamic

PCI Journal | Fa l l 2013

85

Table 3. Bias factors of maximum moment due to live load in 75 years

Span, ft

LL
ADTT250

ADTT1000

ADTT2500

ADTT5000

ADTT10,000

200

1.34

1.36

1.37

1.40

1.40

190

1.35

1.37

1.38

1.41

1.41

180

1.35

1.37

1.39

1.42

1.42

170

1.36

1.38

1.40

1.42

1.43

160

1.37

1.39

1.41

1.43

1.44

150

1.37

1.39

1.41

1.44

1.44

140

1.38

1.40

1.42

1.45

1.45

130

1.38

1.40

1.43

1.45

1.46

120

1.39

1.41

1.44

1.46

1.47

110

1.40

1.41

1.44

1.46

1.47

100

1.40

1.42

1.43

1.45

1.47

90

1.41

1.42

1.43

1.45

1.47

80

1.41

1.42

1.43

1.45

1.47

Source: Data from Rakoczy (2011). Note: ADTT = average daily truck traffic; LL = bias factor of live load. 1ft = 0.305m.

load. Environmental conditions (wind, ice, and temperature) and extreme events (collisions) may have additional
load effects. In this study only primary load components
were considered.
Dead load is the permanent weight of structural and nonstructural members of a bridge. It is convenient to separate
precast concrete elements, cast-in-place components (slab),
and wearing surface (concrete or asphalt). In this study,
the bias factor value of cast-in-place members is 1.05
and coefficient of variation V is 0.10, whereas for precast
concrete components, bias factor is 1.03 and coefficient
of variation V is 0.08. For wearing surface, is 1.00 and
coefficient of variation V is 0.25. The statistical parameters
for dead load were taken from the literature.2,11
The most recent live-load model was developed by Rakoczy3 and Nowak et al.4 based on an extensive weigh-inmotion survey including more than 65million vehicles at
32 different locations. Table3 lists the statistical parameters. The values for 90, 120, and 200ft (27, 36, and 61m)
were taken from research by Rakoczy3 and Nowak et al.4
For other span lengths, the results were linearly interpolated or extrapolated.
The parameters were derived for static load only. In the total
load model, the dynamic load had to be included, and it was
taken as 0.1 of the static load. The coefficient of variation V of
total live load is 0.18 for all span length, as recommended in
National Cooperative Highway Research Program report368.2

86

Fall 2 0 1 3 | PCI Journal

The statistical parameters of the total load also depend

on the dead loadtolive load ratio. For longer bridges,
dead load can be larger than live load, and for short-span
bridges, they can be about the same. Therefore, the reliability analysis was performed for span lengths ranging
from 80 to 200ft (24 to 61m).
In the AASHTO LRFD specifications, the design load
HL-93 is either a three-axle design truck and uniform
load equal to 0.64kip/ft (9.3kN/m) or a design tandem with two axles of 25kip (110kN) spaced at 4ft
(1.2m) and a uniform load of 0.64kip/ft (9.34kN/m).
The load case that produces the greater load effect
governs.
The design loads are multiplied by load factors to obtain a
factored total design load Qd as is presented in Eq.(3).

Qd = 1.75(LL + IM) + 1.25DC + 1.5DW (3)

where
DC = dead load of structural components and nonstructural attachments
DW = dead load of wearing surface and utilities
LL = vehicular live load
IM = vehicular dynamic load allowance

Reliability index

ADTT 250
ADTT 1000
ADTT 2500
ADTT 5000
ADTT 10,000

0
0

20

40

60

80

100

120

140

160

180

200

220

Span, ft
Figure 5. Reliability index for moment versus span with resistance factor = 1.0. Note: ADTT = average daily truck traffic. 1ft = 0.305m.

Reliability analysis

Moreover, the reliability index is related to the probability of failure (Eq.[6]).

The reliability analysis was used in this study for evaluating the structural performance in terms of a reliability
index , defined as a function of the probability of failure.8
The relia bility analysis was performed for the strength
limit state functions formulated for the considered structural types and load components. In the reliability analysis,
load and resistance were treated as random variables. The
statistical parameters for resistance of prestressed concrete
beams were developed in this study based on the new data.
The statistical parameters of load were taken from the most
recent study based on the weigh-in-motion truck data.3,4

where

= standard normal distributions function

The reliability index for a limit state function defined
in Eq.(4) can be calculated using the general formula
Eq.(7).8

Formulation of the limit state function requires a definition

of failure representing a boundary between the desired and
undesired performance of a structural component. The general form of the limit state function is expressed by Eq.(4).

g(R,Q) = R Q (4)

where
g = limit state function
Q = load effect (demand)
If function g is greater than or equal to 0, the structure is
safe (structural capacity is greater than load effect); if g is
less than 0, the structure fails (load effect is greater than
structural capacity). Hence, the probability of failure Pf is
equal to the probability that the limit state function reaches
a negative value (Eq.[5]).

Pf = P(R Q < 0) = P(g < 0)

(5)

(6)

(7)

where
Q = mean load effect

Q = standard deviation of load effect

One of the first steps in the reliability procedure is to
define the limit state function for the considered structural
type. In this study, the flexural limit state of prestressed
concrete girders was considered. The reliability index was
calculated for various average daily truck traffic values
from 250 to 10,000 and for span lengths from 80 to 200ft
(24 to 61m). For each considered span length and girder
spacing, dead load and live load were estimated and treated
as constants. The design of prestressed concrete girders
was according to AASHTO LRFD specifications. Then
the statistical parameters of load, mean load effect Q, and
coefficient of variation for load effect VQ were calculated
using the load model parameters.

PCI Journal | Fa l l 2013

87

0.8
0.6
0.4
0.2
0.0
-0.2

10

20

30

40

50

-0.4
-0.6

Change from nominal value, %

fpsAps

IM

LL

DL

Ratio of girder reliability index

Ratio of girder reliability index

1.0

1.0

0.8
0.6
0.4
0.2
0.0
-0.2

10

20

30

40

50

-0.4
-0.6

Change from nominal value, %

fpsAps

NU 2000 girder

IM

LL

DL

NU 1600 girder

Figure 6. Sensitivity function for NU2000 girder, 200ft (61m) and NU1600 girder, 160ft (49m). Note: Aps = area of prestressing steel; b = width of the compression face of the member; d = effective depth; DL = total dead load (DC + DW); = specified compressive strength of concrete at 28 days; fps = average stress in
prestressing steel at nominal bending resistance; IM = vehicular dynamic load allowance; LL = vehicular live load allowance.

Figure5 plots the results of the reliability analysis for

bending moment with a resistance factor of 1.0 versus
span length. The results show that is about 3.6 for lower
average daily truck traffic and 3.4 for higher average daily
truck traffic. Also, shorter bridges have a slightly higher
than longer span bridges.

Sensitivity analysis
The performance of the structure can be underestimated
or overestimated because of insufficient data, time-related
changes, or human error. Some parameters have a major
effect on overall performance, while others do not. The important parameters can be identified by sensitivity analysis.
Therefore, sensitivity analysis was conducted for typical
prestressed, precast concrete girders. The considered load,
resistance, and section geometry parameters were dead
load DL (sum of DC and DW), live load LL, dynamic load
allowance IM, material strengths
and fps, effective depth
d, width of the compression face of the member b, and area
of prestressing steel Aps.
Load components were treated as separate parameters
and were increased one at a time by 10% from the nominal value. Resistance parameters were grouped in three
subgroups: fpsAps, b , and d. To check the sensitivity of resistance parameters, the reliability indices were calculated
for their nominal values and then for resistance parameters
reduced by multiples of 10%.
The reliability indices were calculated for the AASHTO
and NU concrete girders with spans from 100 to 200ft (30
to 61m). Figures6 and 7 show the drop of the reliability
index for reduced/increased load and resistance parameters
for all of the considered girders. The vertical axis is the

88

Fall 2 0 1 3 | PCI Journal

ratio of the reliability index corresponding to either an

increased load component or reduced resistance parameter
and the reliability index for an intact girder.
The results indicate that reliability depends mostly on the
strength fps and cross-sectional area Aps of the prestressing steel and effective depth d. The reduction of effective
depth by 30% to 35% reduced the reliability index to the
negative values. The concrete strength , effective width
b, and dynamic load allowance IM are not important.
The reliability of a girder depends mostly on the resistance parameters. Reduction of the critical parameters,
that is, fps and Aps, reduces the reliability index (Fig.8) for
the considered types of girders. The vertical axis represents the ratio of the reliability index corresponding to
reduced resistance parameter and the reliability index for
an intact girder. The reliability index drops by 100% when
prestressing steel fpsAps is reduced about 40%. Similarly,
Fig.8 shows the effect of a reduced effective depth d. This
parameter is more important because, for example, a decrease of effective depth d by 30% to 35% causes a 100%
decrease in the reliability index.

Conclusion
The statistical parameters of resistance were considered
for AASHTO and NU girders designed according to the
AASHTO LRFD specifications. Based on a large material test data set for prestressing strands provided by the
industry, new statistical parameters of resistance were
derived. The results indicate that concrete and reinforcing
steel properties have improved over the past 30years,9 and
this can have a positive effect on the load-carrying capacity
of structural components.

0.8
0.6
0.4
0.2
0.0
-0.2

10

20

30

40

50

-0.4
-0.6

Ratio of girder reliability index

Ratio of girder reliability index

1.0

IM

0.8
0.6
0.4
0.2
0.0
-0.2

LL

10

20

30

40

50

LL

DL

-0.4
-0.6

Change from nominal value, %

fpsAps

1.0

Change from nominal value, %

fpsAps

DL

AASHTO Type IV girder

IM

AASHTO Type III girder

Figure 7. Sensitivity functions for AASHTO TypeIV girder, 120ft (36m) and AASHTO TypeIII girder, 100ft (30m). Note: Aps = area of prestressing steel; b = width of
the compression face of the member; d = effective depth; DL = total dead load (DC + DW); = specified compressive strength of concrete at 28 days; fps = average
stress in prestressing steel at nominal bending resistance; IM = vehicular dynamic load allowance; LL = vehicular live load allowance.

On the other hand, the new statistics for live load indicate
that traffic volumes and truck weights have increased.3,4
Increased loading has a negative effect on the reliability
index.
The reliability analysis was performed for AASHTO and
NU prestressed, precast concrete girders with spans from
80 to 200ft (24 to 61m) and average daily truck traffic
from 250 to 10,000 vehicles using the new statistics for
resistance and load effect.
The sensitivity analysis was performed for typical prestressed concrete girders. The considered parameters
included load components, material strength, and section
geometry. The results of sensitivity analysis indicate that

the reliability is mostly affected by the prestressing steel

fpsAps and effective depth d. The reduction of effective
depth by 30% to 35% caused a 100% decrease in the
reliability index. Reliability indices drop 100% when
prestressing steel fpsAps is reduced about 40%. The least important parameters are concrete strength b and dynamic
load allowance IM.

References
1. Nowak, A. S., A. S. Yamani, and S. W. Tabsh. 1994.
Probabilistic Models for Resistance of Concrete Bridge
Girders. ACI Structural Journal 91 (3): 269276.
2. Nowak, A. S. 1999. Calibration of LRFD Bridge De-

1.0

0.8
0.6
0.4
0.2
0.0
-0.2

10

20

30

40

50

-0.4
-0.6

Change from nominal value, %

Ratio of girder reliability index

Ratio of girder reliability index

1.0

0.8
0.6
0.4
0.2
0.0
-0.2

10

20

30

40

50

-0.4
-0.6

Change from nominal value, %

NU 2000

NU 1800

NU 1600

NU 2000

NU 1800

NU 1600

NU 1350

AASHTO IV

AASHTO III

NU 1350

AASHTO IV

AASHTO III

Effective depth d

Figure 8. Effect of decrease in prestressing steel fpsAps and effective depth d for prestressed concrete girders. Note: 1ft = 0.305m.
PCI Journal | Fa l l 2013

89

sign Code. National Cooperative Highway Research

Project report 368. Washington, DC: TRB (Transportation Research Board), National Academy Press.
3. Rakoczy, P. 2011. WIM Based Load Models for
Bridge Serviceability Limit States. PhD diss., University of NebraskaLincoln.
4. Nowak, A. S., P. Rakoczy, and J. M. Kulicki. 2012.
WIM-Based Live Load for Bridges. In TRB 91st
Annual Meeting Compendium of Papers DVD, paper
12-4109. Washington, DC: TRB.
5. AASHTO (American Association of State Highway
and Transportation Officials). 2011. AASHTO LRFD
Bridge Design Specifications. 6th ed. Washington, DC:
AASHTO.
6. NDOR (Nebraska Department of Roads) Bridge Division. 2009. NDOR Bridge Office Policies and Procedures. Lincoln, NE: NDOR.
7. Morcous, G., and A. Akhnoukh. 2007. Reliability
Analysis of NU Girders Designed Using AASHTO
LRFD. In Proceedings of the 2007 Structures
Congress, Long Beach, CA, May 2007. Reston, VA:
American Society of Civil Engineers.
8. Nowak, A. S., and K. R. Collins. 2012. Reliability of
Structures. 2nd ed. New York, N.Y.: CRC Press.
9. Nowak, A. S., A. M. Rakoczy, and E. Szeliga. 2012.
Revised Statistical Resistance Models for R/C
Structural Components. In Andy Scanlon Symposium
on Serviceability and Safety of Concrete Structures:
From Research to Practice, ACI SP-284-6, pp. 116.
Farmington Hills, MI: ACI (American Concrete
Institute).
10. Rakoczy, A. M., and A. S. Nowak. 2013. Resistance
Model of Lightweight Concrete Members. ACI Materials Journal 110 (01): 99108.
11. Ellingwood, B., T. V. Galambos, J. G. McGregor, and
C. A. Cornell. 1980. Development of a Probability
Based Load Criterion for American National Standard
A58. NBS (National Bureau of Standards) special
report 577. Washington, DC: U.S. Department of
Commerce NBS.
12. Nowak, A. S., and A. M. Rakoczy. 2012. Reliability
of Prestressed Concrete Girders. In PCI Convention
and National Bridge Conference, Nashville, TN, September 29October 2, 2012. Chicago, IL: PCI.

90

Fall 2 0 1 3 | PCI Journal

Notation
a = depth of the equivalent stress block = c1
Aps = area of prestressing steel
As = area of mild steel tensile reinforcement
= area of mild steel compressive reinforcement
b = width of the compression face of the member; for a
flange section in compression, the effective width of
the flange
bw = web width or diameter of a circular section
c

= distance from extreme compression fiber to neutral

axis

d = effective depth
dp = distance from extreme compression fiber to the centroid of prestressing tendons
ds = distance from extreme compression fiber to the centroid of mild steel tensile reinforcement
= distance from extreme compressive fiber to the centroid of mild steel compressive reinforcement
DC = dead load of structural components and nonstructural attachments
DL = total dead load (DC + DW)
DW = dead load of wearing surfaces and utilities
= specified compressive strength of concrete at 28
days, unless another age is specified
fps = average stress in prestressing steel at nominal bending resistance
fs = stress in mild steel tensile reinforcement at nominal
flexural resistance
= stress in mild steel compressive reinforcement at
nominal flexural resistance
fu = ultimate strength of reinforcing steel
fy = yield stress of reinforcing steel
F = fabrication factor

g = limit state function

R = standard deviation of resistance

hf = compression flange depth of an I or T member

= resistance factors

IM = vehicular dynamic load allowance

= standard normal distribution function

LL = vehicular live load allowance

M = material factor
P = professional factor
Pf = the probability of failure
R = resistance (capacity)
Rn = nominal value
Q = load effect (demand)
Qd = factored total design load
V = coefficient of variation
VF = coefficient of variation for fabrication factor F
VP = coefficient of variation for professional factor P
VQ = coefficient of variation for load effect Q
VR = coefficient of variation of resistance R

= reliability index, defined as a function of the probability of failure

1 = stress block factor
= bias factor
F = bias factor for fabrication factor F
LL = bias factor of live load
P = bias factor for professional factor P
R = bias factor of resistance R
= mean value
Q = mean load effect
R = mean resistance

= standard deviation
Q = standard deviation of load effect

PCI Journal | Fa l l 2013

91

About the authors

Anna M. Rakoczy is a postdoctoral research assistant in the
Department of Civil Engineering
at the University of Nebraska
Lincoln. She received her PhD
from UNL in 2012. Her major
research area is structural engineering, and she has studied and worked in reliability
and risk analysis, structural analysis, and design code
calibration. Her research interests include the development of reliability-based design criteria for lightweight
concrete structures and evaluation criteria for serviceability limit states of railway bridges.
Andrzej S. Nowak has been a
professor of civil engineering at
the University of NebraskaLincoln since 2005, after 25 years at
the University of Michigan. He is
a Fellow of ACI and member of
ACI committees 343, Concrete
Bridges and 348, Structural Safety. His major contribution is the development of reliability-based calibration
procedure for limit state design codes. The procedure
was successfully applied to the development of the
AASHTO LRFD code and calibration of resistance
factors for ACI 318.

The objective of this paper is to present the results of

the reliability analysis for prestressed concrete girders
using the most recent live load and resistance (flexure) models. Several types of American Association
of State Highway and Transportation Officials girders
and Nebraska University (NU) girders are considered
with spans ranging from 80 to 200ft (24 to 61m) and
girder spacings from 8 to 10ft (2.4 to 3m).
An important part of the research presented is derivation of sensitivity functions for various parameters that
affect performance of prestressed concrete girders. The
focus is on strength limit states, in particular bending
capacity. The contribution of several resistance parameters, such as reinforcement area and yield stress, is
considered. The procedure is demonstrated on representative girder bridges.

Keywords
Girder, load and resistance factor design, LRFD, reliability analysis, sensitivity analysis, statistical parameters, strand.

Abstract

Review policy

Structural performance of bridges depends on the

applied loads and the load-carrying capacity, both of
which are random in nature. Therefore, probabilistic
methods are needed to quantify safety. It has been
widely agreed to measure structural reliability in terms
of a reliability index.

This paper was reviewed in accordance with the

Precast/Prestressed Concrete Institutes peer-review
process.

Material properties have changed over the last 30

years. The new material test data include compressive strength of concrete, yield strength of reinforc-

92

ing steel, and tensile strength of prestressing strands.

Average daily truck traffic and gross vehicle weight
have also changed. The updated statistical parameters
for prestressed concrete girders and live load affect the
reliability indices.

Fall 2 0 1 3 | PCI Journal

Reader comments
Please address and reader comments to journal@pci
.org or Precast/Prestressed Concrete Institute, c/o PCI
Journal, 200 W. Adams St., Suite 2100, Chicago, IL
60606. J