CALIBRATION OF LRFD BRIDGE CODE

Andrzej S. Nowak, l Member, ASCE

ABSTRACT: This paper reviews the code development procedures used for the new load and resistance factor
design (LRFD) bridge code. The new code is based on a probability-based approach. Structural performance
is measured in terms of the reliability (or probability of failure). Load and resistance factors are derived so
that the reliability of bridges designed using the proposed provisions will be at the predefined target level.
The paper describes the calibration procedure (calculation of load and resistance factors). A new live load
model is prop.osed, which provides a consistent safety margin for a wide spectrum of spans. The dynamic load
model takes mto account the effect of road roughness, bridge dynamics, and vehicle dynamics. Statistical
models of resistance (load-carrying capacity) are summarized for noncomposite steel, composite steel, rein-
forced concrete, and prestressed concrete. The reliability indices for bridges designed using the proposed code
are compared with the reliability indices corresponding to the current specification. The proposed code pro-
Downloaded from ascelibrary.org by Auburn University on 05/31/16. Copyright ASCE. For personal use only; all rights reserved.

visions allow for a consistent design with a uniform level of reliability.

INTRODUCTION described in other papers (Nowak 1993; Nowak and Hong

1991; Hwang and Nowak 1991; Tabsh and Nowak 1991; Ting
The objective of this paper is to present the procedures and Nowak 1991; Nowak et al. 1993). Load and resistance
used in the calibration of a new load and resistance factor are treated as random variables and are described by bias
design (LRFD) bridge code. The allowable stress method and factors (ratio of mean to nominal), denoted by A, and by
load factor design, specified in the current AASHTO code coefficients of variation, denoted by V.
(Standard 1992), do not provide for a consistent and uniform
safety level for various groups of bridges. One of the major
goals of the new code is to provide a uniform safety reserve. CALIBRATION PROCEDURE
The main parts of the current AASHTO (Standard 1992) The development of a new code involves the following
specification were written about 50 yr ago. There were many steps:
changes and adjustments at different times, which resulted
in gaps and inconsistencies. Therefore, the work on the LRFD
code also involves rewriting the document based on the state- 1. Selection of representative bridges: About 200 struc-
of-the-art knowledge about various branches of bridge en- tures were selected from various geographical regions
gineering. This paper summarizes some of these changes re- of the United States. These structures cover materials,
lated to load and resistance models. types, and spans, which are characteristic of the region.
The theory of code writing has advanced in the last 20 yr. Emphasis is placed on current and future trends; instead
Some of the important contributions were summarized by of on very old bridges. For each selected bridge, load
Madsen et al. (1986), Melchers (1987), Ellingwood et al. (1980), effects were calculated for various components. Load-
and Nowak and Lind (1979). The major tool in the devel- carrying capacities were also evaluated.
opment of a new code is the reliability analysis procedure. 2. Establishment of statistical database for load and resis-
Structural performance is measured in terms of the reliability tance parameters: The available data on load compo-
or probability of failure. The code provisions are formulated nents, including results of surveys and other measure-
so that structures designed using the code have a consistent ments, were gathered. Truck survey and weigh-in-motion
and uniform safety level. The available reliability methods (WIM) data were used for modeling live load. There
are reviewed in several textbooks (Thoft-Christensen and Baker was little field data on the dynamic load and, therefore,
1982; Madsen et al. 1986; Melchers 1987). The methods vary a numerical procedure was developed to simulate the
with regard to accuracy, required input data, computational dynamic bridge behavior. Statistical data for resistance
effort, and special features (time variance). include material tests, component tests, and field mea-
In an LRFD code, the basic design formula is surements. Numerical procedures were developed to
simulate the behavior of large structural components
L 'V)( < </>R n (1) and systems.
where Xi = nominal (design) load component i; 'Vi load 3. Development of load and resistance models: Loads and
resistance are treated as random variables. Their vari-
factor i; R n = nominal (design) resistance; and </> = resistance
factor. The objective of calibration is to determine load and ation is described by cumulative distribution functions
resistance factors so that the safety of bridges designed ac- (CDF) and correlations. The CDFs for loads were de-
cording to the code will be at the preselected target level. rived using the available statistical database (step 2).
This paper presents the calibration procedure, including The live load model includes the multiple presence of
load models, resistance models, reliability analysis, and the trucks in one lane and in adjacent lanes. Multilane re-
development of load and resistance factors. Bridge load and duction factors were calculated for wider bridges. The
resistance models are only summarized here because they are dynamic load was modeled for single trucks and two
trucks, side-by-side. Resistance models were developed
'Prof., Dept. of Civ. Engrg., Univ. of Michigan, Ann Arbor, MI for girder bridges. The variation of the ultimate strength
48109-2125. was determined by simulations.
Note. Associate Editor: Dennis R. Mertz. Discussion open until Jan- 4. Development of reliability analysis procedure: Struc-
uary 1, 1996. To extend the closing date one month, a written request tural performance is measured in terms of the reliability
must be filed with t~e ASCE Manager of Journals. The manuscript for
thiS paper was submitted for review and possible publication on January or probability of failure. Reliability is measured in terms
8, 1993. This paper is part of the Journal of Structural Engineering, Vol. of the reliability index 13, calculated by using an iterative
121, No.8, August, 1995. ©ASCE, ISSN 0733-9445/95/0008-1245-1251/ procedure. The developed load and resistance models
$2.00 + $.25 per page. Paper No. 5403. (step 3) are part of the reliability analysis procedure.
JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 1995/1245

J. Struct. Eng., 1995, 121(8): 1245-1251

 5. Selection of target reliability index: Reliability indices laJ Standard HS20 Truck
were calculated for a wide spectrum of bridges designed 142.3 kN 142.3 kN
according to the current AASHTO report ("Standard"
1992). The performance of existing bridges was evalu-
ated to determine whether their reliability level is ad-
equate. The target reliability index 131' was selected to
____* 35.6kN

*
I I
t _
provide a consistent and uniform safety margin for all
structures.
6. Calculation of load and resistance factors: Load factors (hI HS20 Lane Loading
'Yare calculated so that the factored load has a prede-
termined probability of being exceeded. The relation- 80 kN (for moment)
ship among nominal (design) load X m mean load m x , 115 kN (for shear)
and factored load 'YiXn is shown in Fig. 1. The corre- 9.3 kN/m
sponding terms for resistance are shown in Fig. 2. Re-
sistance factors, <1>, are calculated so that the structural
reliability is close to the target value 131"
Downloaded from ascelibrary.org by Auburn University on 05/31/16. Copyright ASCE. For personal use only; all rights reserved.

leI Military Loading

LOAD MODELS
Load models are an important part of the code. The major
components of bridge loads are dead load, live load, dynamic _____ 10_7_kN_L~_10_7_kN _

load (impact), environmental loads (e.g., wind, earthquake,

temperature, water pressure), and special loads (e.g., colli- 1~·2m.1
sion forces). The code specifies design loads, load combi- FIG. 3. Nominal Live Load; HS-20 Truck and Lane Load (Standard
nations, and load factors. However, the calculation of load 1992)
factors requires a knowledge of the statistical models, in par-
ticular the distribution of magnitude, rate of occurrence, time
variation, and correlation with other load components. The
development of the live load model for bridges is presented 2
in another paper (Nowak 1993). Therefore, it is only sum-
marized in this paper with the resulting statistical parameters.
Dead load D is the weight of structural and nonstructural
members and is calculated using specified densities of ma- Shear
terials. In the reliability analysis, several categories of Dare
considered, because of differences in the degree of variation.
For factory-made components (structural steel, precast con-
crete) the bias factor A is 1.03 and the coefficient of variation
V is 0.08; for cast-in-place concrete X. is 1.05 and V is 0.10;
and for asphalt surface the mean thickness is equal to 75 mm
(3 in.) and Vis 0.25 (Nowak 1993). In the current AASHTO
code (Standard 1992), a dead load factor of 1.3 is applied to o-l--.----,.-...--,-~-r__~__r-~-r-...,..._j

all components of dead load. o 10 20 30 40 50 60

The current design live load is based on a HS-20 truck, Span (m)
lane loading, or military loading, as shown in Fig. 3 (Standard
FIG. 4. Bias Factors for Moments and Shears; AASHTO (Standard
1992). The corresponding moments and shears are lower than 1992)
the actual load effects of heavy traffic observed on the high-
ways in the United States today. The actual moments and
shears caused by the heaviest vehicles observed in truck sur-
veys range from 1.5 to 1.8 times the design moments and
shears (calculated using the HS-20 load). In the United States
fleW an average lifetime for bridges is about 75 yr. Therefore, this
time period is used as the basis for the calculation of loads.
Using the available data, a statistical model was developed
for the mean maximum 75-yr moments and shears by ex-
trapolation of the truck survey data. For a single lane, the
o mx bias factors (ratios of the mean maximum 75-yr moment/shear
and HS-20 moment/shear) are plotted versus the span length
FIG. 1. Nominal Load, Mean Load, and Factored Load
in Fig. 4.
Various new design loads were considered to obtain a uni-
form bias factor. Good results were achieved by superposition
of the HS-20 truck and a uniformly distributed load of 9.3
kN . m (640 lb/ft). A tandem is specified for shorter spans.
The proposed live load is shown in Fig. 5. The bias factors
for moments and shears calculated for various spans using
the new load are presented in Fig. 6.
o R
For multiple lanes, the girder moment or shear is a result
of more than one truck load. It was determined by simulations
FIG. 2. Nominal Resistance, Mean Resistance, and Factored Re- that for two-lane bridges, two side-by-side trucks govern, with
sistance fully correlated weights. The probability of having two very
1246/ JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 1995

J. Struct. Eng., 1995, 121(8): 1245-1251

 (a) Truck and Uniform Load TABLE 1. Proposed Design Multilane Factors
Number of Lanes
142.3kN 142.3kN
ADTT One Two Three Four or more
(1) (2) (3) (4) (5)
9.3kN/m

.
100 1.15 0.95 0.65 0.55
1,000 1.20 1.00 0.85 0.60
I 4.27m
~.
I 4.27-9.14m
~
I 5,000 1.25 1.05 0.90 0.65

--
(h) Tandem and Uniform Load

Span:
12m

I - -<>-- 18m

++
1l1.2kN Ill1.2kN 2 27m

9.3kN/m
-
--+-- 36m
60m
Downloaded from ascelibrary.org by Auburn University on 05/31/16. Copyright ASCE. For personal use only; all rights reserved.

IJ·2m~1
FIG. 5. Proposed Nominal Live Load (LRFD Load)

2 -- AASlITO (1992)

O+---.----r--.--,....-~--r-~--l

.... ~
o 2 3 4
Shear Girder Spacing (m)

FIG. 7. GDFs Specified by AASHTO and Proposed LRFD Code

GDF = 0.4 + (s/6) - (s/25)2 (4)

where Span = span length in ft (1 ft = 0.305 m). For com-
o-l--.---r-~-.,.---,.-.---.---.-..---.----.-----l parison, GDFs calculated using (3) and (2) are shown in Fig. 7.
o 10 20 30 40 50 60 Dynamic load is defined as the ratio of dynamic deflection
Span (m) and static deflection. The current AASHTO (Standard 1992)
report specifies impact I as a function of span length only
FIG. 6. Bias Factors for Moments and Shears; Proposed LRFD
Load I = 50/(125 + Span) (5)

heavy trucks simultaneously on the bridge is lower than hav- where Span = span length in ft (1 ft = 0.305 m). The actual
ing a single heavily loaded truck. It was calculated that in a dynamic load depends on three major factors: road rough-
side-by-side occurrence, each truck is about 85% of the mean ness, bridge dynamics (natural period of vibration), and ve-
maximum 75-yr truck, which corresponds to the rriean max- hicle dynamics (type and condition of suspension system).
imum 2-month truck. Traffic frequency is very important in The derivation of the statistical model for the dynamic be-
the statistical analysis of the heaviest trucks. Average daily havior of bridges is presented by Hwang and Nowak (1991).
truck traffic (ADTT) varies depending on local conditions. The simulations indicated that the mean dynamic load is less
The calculations were performed for ADTT equal to 100, than 0.17 for a single truck and less than 0.12 for two trucks,
1,000 and 5,000 trucks (in one direction). The proposed mul- for all spans. The coefficient of variation of the dynamic load
tilane factors are presented in Table 1. is 0.80. The coefficient of variation of a joint effect of the
The total moment is distributed to girders. The girder mo- live load and dynamic load is 0.18.
ment can be determined by using a girder distribution factor The proposed new design dynamic load is 0.33, applied to
(GDF). In the present AASHTO code (Standard 1992), GDFs the truck effect only, with no dynamic load applied to the
for moments are specified only as a function of girder spacing, uniformly distributed portion of the live load (Fig. 5).
s
RESISTANCE MODELS
GDF = sid (2)
The capacity of a bridge depends on the resistance of its
where s = girder spacing in ft (1 ft = 0.305 m); and d = 5.5 components and connections. The component resistance R is
for steel girders and prestressed concrete girders, and 6.0 for mostly determined by materials strength and dimensions. R
reinforced-concrete T-beams. GDF is applied to half of the is treated as a random variable. The causes for uncertainty
lane moment. For shear, the GDF in (2) is specified, except can be put into the following categories:
for the axle of the HS-20 truck placed directly over the sup-
port. New GDFs were derived by Zokaie et al. (1991), which 1. Material: strength of material, modulus of elasticity,
relate GDF to girder spacing and span length. For the mo-
cracking stress, and chemical composition.
ment 2. Fabrication: geometry, dimensions, and section mod-
GDF = 0.15 + (s/3)OIi(s/Span)02 (3) ulus.
3. Analysis: approximate method of analysis, and idealized
and for shear stress and strain distribution model.
JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 1995/1247

J. Struct. Eng., 1995, 121(8): 1245-1251

 The resulting variation of resistance has been modeled by where <1>-1 = inverse standard normal distribution function.
tests, observations of existing structures, and by numerical Examples of {3s and corresponding PfoS are shown in Table 3.
simulations. In this study, R is considered a product of the The available procedures to calculate {3 vary with regard
nominal resistance R and three parameters: strength of ma-
II to accuracy, required input data, and computing effort. In
terial M, fabrication (dimensions) factor F, and analysis this study, the reliability analysis is performed using an it-
(professional) factor P, as was suggested by Ravindra and erative method based on normal approximations to non nor-
Galambos (1978) mal distributions at the so-called design point [Thoft-Chris-
tensen and Baker (1992), Melchers (1987)]. The design point
R = R"MFP (6) is the point of the maximum probability (maximum joint
The mean value of R is a product of the mean values of probability density of Rand Q) on the failure boundary (limit
M, F, and P, and the coefficient of variation V R is state function). The mathematical representation of the fail-
ure boundary is the limit state function equal to zero, g = R
(7) - Q = O. The design point, denoted by (R*, Q*), is located
on the failure boundary, so R* = Q*.
where V M, Vh and VI' = coefficients of variation of M, F, Let F R be the cumulative distribution function (CDF) and
and P, respectively. fR the probability density function (PDF) for R. Similarly, F(}
Downloaded from ascelibrary.org by Auburn University on 05/31/16. Copyright ASCE. For personal use only; all rights reserved.

For this calibration, the resistance parameters are based and fQ are the CDF and PDF for Q. An initial value of R*
on the available material and component tests. Flexural ca- (design point) is guessed first. Next, F R is approximated by
pacity is established by simulation of the moment-curvature a normal distribution F~, such that
relationship, as described by Tabsh and Nowak (1991). The
shear capacity of concrete components is calculated using the (II, 12)
modified compression field theory (Collins and Mitchell 1991). The standard deviation and the mean of R' are
The statistical parameters were developed by Nowak et al.
(1993). rr~ = 4>,,{<I>-I[FR(R*)]}/fR(R*) (13)
The models of resistance are considered for noncomposite
m~ = R* - rr;,<I> I[FR(R*)] ( 14)
steel girders, composite steel girders, reinforced-concrete T-
beams, and prestressed-concrete AASHTO-type girders. where 4>11 = PDF of the standard normal random variable;
Nominal (design) value of resistance is calculated using (1). and <I> = CDF of the standard normal random variable.
The bias factors and coefficients of variation for the consid- Similarly, F Q is approximated by a normal distribution
ered materials and limit states are summarized in Table 2. F Q, such that

RELIABILITY ANALYSIS F~(Q*) = FQ(Q*); f~(Q*) = fQ(Q*) (15. 16)

The available reliability methods are presented in several The standard deviation and mean of Q' are
publications [e.g., Thoft-Christensen and Baker (1982), Mad-
rr~ = 4>,,{<I>-I[FQ(Q*)]}/fQ(Q*) (17)
sen et al. (1986)]. In this study the reliability analysis is per-
formed using an iterative procedure. m~ = Q* - rr~<I>-I[FQ(Q*)] (18)
Let R represent the resistance (e.g., moment carrying ca-
pacity) and Q represent the load effect (e.g., total moment The reliability index is
applied to the considered beam). Then, the corresponding
13 = (m~ - m~)/(rr;l + rrJ)1I2 (19)
limit state function g can be written as
Next, a new design point can be calculated from the fol-
g=R-Q (8) lowing equations:
If g > 0 the structure is safe, otherwise it fails. The probability (20)
of failure P r is equal to
(21 )
PF = Prob(R - Q < 0) = Prob(g < 0) (9)
Then, the second iteration begins; the approximating nor-
It is convenient to measure structural safety in terms of a mal distributions are found for F R and F Q at the new design
reliability index {3, defined as a function of P r point. The reliability index is calculated using (19), and the
next design point is found from (20) and (21). Calculations
(10) are continued until R* and Q* do not change in consecutive
iterations.
TABLE 2. Statistical Parameters of Resistance Resistance is a product of parameters M, F, and P; there-
Coefficient of fore, it is assumed that the cumulative distribution function
Type of structure Bias factor variation of R is lognormal. The CDF of the load is treated as a normal
(1 ) (2) (3) distribution function because Q is a sum of the components
Noncomposite steel girder
Moment 1.11 0.115 TABLE 3. Reliability Index and Probability of Failure
Shear 1.14 0.12 Probability of failure
Reliability index
Composite steel girders
(1 ) (2)
Moment 1.11 0.12
Shear 1.14 0.12 o 0.5
Reinforced-concrete T-beams 1 0.159
Moment 1.14 0.13 2 (1.0228
Shear 1.165 0.16 3 0.00135
Prestressed-concrete girders 4 (J. 0000317
Moment 1.05 0.Q75 5 0.000000287
Shear 1.165 0.16 6 0.000000000987

1248/ JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 1995

J. Struct. Eng., 1995, 121(8): 1245-1251

 5-ro--------------, s = girder spacing
5 ....-------------,
s • girder spacing

s -3.6m j 4 s.3.6m

-
s =3.0m

s=2.4m
f
.l!
3
2
-----
s .3.0m

s = 2.4 m

~ 8=I.am ~ s-1.8m
s. 1.2 m s = 1.2 m

0.1-------------.
o 50
.. 0.1----~---------'
o 50
Span (JIll
Span (m)
FIG. 8. 13s for Steel Girders-Moments; AASHTO (Standard 1992)
FIG. 12. 13s for Reinforced-Concrete T-Beams-Shears; AASHTO
(Standard 1992)
5....---------------, s = girder spacing
Downloaded from ascelibrary.org by Auburn University on 05/31/16. Copyright ASCE. For personal use only; all rights reserved.

s·3.6m 5.,.------------, s • girder spadng

s .3.0m
s.3.6m
- . - - s=2.4m
s -3.0m
2
- s-1.8m
-- s=2.4m
s - 1.2 m
--0-- sal.am
O.L..-~-~----------' s - 1.2 m
o 50
Span (JIll O.L..-~------------'
o 50
FIG. 9. 13S for Reinforced-Concrete T-Beams-Moments; AASHTO
(Standard 1992)
Span (m)

FIG. 13. 13s for Prestressed-Concrete Girders-Shears; AASHTO

5....--------------, s = girder spacing (Standard 1992)
s=3.6m
2
s=3.0m

.-.-- s = 2.4 m
.....
- s=1.8m
s. 1.2 m

O.!--~--------~---J
o 50
--
---
---
o (factory-made component')
o (cast-ln-pJace concrete)
o 'asphalll
L Uncludlrlg dynamic loadl
Span (JIl) o
1.5 2.0 2.5
RG. 10. 13s for Prestressed-Concrete Girders-Moments; AASHTO
It
(Standard 1992)
FIG. 14. Load Factors versus k
5....-----------------, s • girder spadng
5~--------------'
s -3.6m

s =3.0m
---+---------------
~---_et'---~---~-----------
s" 2.4 m
-a-- Se 1.8m

s" 1.2 m

----.--- q,-O.95."I"1.70
0.1----~------~-..J
o 50

Span (m)
1
----'"0'-.- 41=0.95."1"1.60
<ll= 1.00. "I" I. 70
<1>=1.00. "1'=1.60

FIG. 11. 13s for Steel Girders-Shears; AASHTO (Standard 1992) 10 20 30 40 50 60

of dead load, live load, and dynamic load. Each design case Span (m)
is defined by a set of nominal loads. The statistical parameters FIG. 15. 13s for Steel Girders-Moments; Proposed LRFD Code
(bias factors and coefficients of variation) are available for
load and resistance. For a given set of load and resistance
The standard deviation (J"~, and mean m~, of the approx-
factors, the nominal resistance is determined from the basic
imating normal distribution of R at R* are
design formula [(1)]. The mean load m Q , standard deviation
of load (J"Q' and mean resistance mR are calculated. The design (23, 24)
point is expressed in terms of a parameter k as follows:
The CDF of the load is a normal distribution function;
(22) therefore, the standard deviation (J" Q = (J" Q and the mean
JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 1995/1249

J. Struct. Eng., 1995, 121(8): 1245-1251

 5 "T"'""---------------. 5

4
~-~
M M
u 4
i ()o---~ _o_- __
:;
t- 3
;"-;-:0::;---1
0-----00--_-0-___

f: 1
---
... •...
_.. -~-_.
.-o.85.r-I.70
.-0.8.5. r-1.6O
.-0.90. r-1.70
41'"0.90. r-1.6O
51
S2
~
1
--..
--- -..
---'9-""
41'"0.90. ,.,.1.70
.-0.90. ,.,.1.60
.-0.95. r-1.70
4looO.95. r-1.6O
0
10 20 30 40 50 60 0 10 20 30 40 50 60

Span (m) Span (m)

FIG. 19. ~s for Reinforced-Concrete T-Beams-Shears; Pro-

Downloaded from ascelibrary.org by Auburn University on 05/31/16. Copyright ASCE. For personal use only; all rights reserved.

FIG. 16. ~s for Reinforced-Concrete T-Beams-Moments; Pro-

posed LRFD Code posed LRFD Code

5-r-----------------. 5.,----------------,
------.,--- ... -----------
0-----0----00-----0------------
-0-----00:::-0-:::::::::::::::
-....-. .-0.95. ,.,.1.70
~ cpO.90.y:=i.70

-
... _ .-0.95. ,.,.1.60
---..-_. lFO.90. y:s 1.60
......,..... ~.95.r-1.70
.. _-~ ... q,ool.OO. ,.,.1.70
q,ool.OO. r-1.6O
---"0-_ 0
4>=0.95. y= 1.60
o +--.......--...-...-....--.......-..--.-~
10 20 30 40 50 60 o 10 20 30 40 50 60

Span (m) Span (m)

FIG. 17. ~s for Prestressed-Concrete Girders-Moments; Pro- FIG. 20. ~s for Prestressed-Concrete Girders-Shears; Pro-
posed LRFD Code posed LRFD Code

5-r---------------, <l>R> 1.3D +2.17(1 + I)L (27)

-
; -o.
....~:::-o------------
------------ where R, D, L, and I are nominal (design) values: L = HS-
20 load, as shown in Fig. 3; I = 50/(125 + Span) where span
= span length in ft (1 ft = 0.305 m); <l> = 1.0 for steel girders;
<l> = 1.0 for prestressed-concrete girders; and <l> = 0.9 for

----.. -.-
......,.....
.. ....
.-0.95. ,.,. •.70
.-0.95. ,.,.1.60
.,.•.00. r-1.70
reinforced-concrete T-beams.
The calculations are carried out for moments and shears.
Several girder spacings s are considered, from 1.2 m to 3.6
m (4-12 ft). The results for moments are shown in Figs. 8-
_-~-
<l>=1.00.r-1.6O 10 for steel girders, reinforced-concrete T-beams, and pre-
stressed-concrete girders for spans of up to 60 m (200 ft). For
10 20 30 40 50 60 shears, the results are presented in Figs. 11-13. There is a
considerable variation in 13-values. In general, reliability in-
Span (m) dices are higher for larger girder spacings. This is an indication
FIG. 18. ~s for Steel Girders-Shears; Proposed LRFD Code that, in most cases, GDFs specified by AASHTO (Standard
1992) are conservative, as shown in Fig. 7. Reliability is the
lowest for a 9-m span (30 ft) and the highest for an 18-m span
mQ = mQ • The reliability index ~, is
(60 ft).
13 = (m~ - mQ)/(a'J + a~)112 (25)
PROPOSED LOAD AND RESISTANCE FACTORS
and the new design point R* is
Preliminary values of load factors for dead load and live
(26) load are determined so that factored loads have the same
probability of being exceeded
It is convenient to start the computations with k 2 in
(22). In most cases, the calculations do not require more than "I = A(1 + kV) (28)
two to three iterations.
where "I = load factor; A = bias factor; V = coefficient of
RELIABILITY INDICES FOR AASHTO (Standard 1992) variation; and k = constant. The relationship between 'Y and
k for dead load components and for a live load is shown in
The reliability indices are calculated for bridge girders de- Fig. 14.
signed using the AASHTO (Standard 1992) specifications. For k = 2, the probability of being exceeded is about 0.02
The design equation is and the corresponding values of load factors are 1.25 for a
1250 I JOURNAL OF STRUCTURAL ENGINEERING I AUGUST 1995

J. Struct. Eng., 1995, 121(8): 1245-1251

 dead load, except 1.5 for asphalt, and 1.6-1.7 for a live load. The reliability analysis is carried out for various possible
For convenience of the code format, the same load factor is values of load and resistance factors. The recommended val-
applied to the static and dynamic portion of the live load. ues provide a uniform safety level for the considered design
Load and resistance factors are rounded to the nearest 0.05. cases.
The considered design formula for the LRFO code is
ACKNOWLEDGMENTS
<l>R> 1.25D + + )'L(1 + I)L
1.5D A (29) The calibration of the new LRFD code was carried out in conjunction
with the National Cooperative Highway Research Program (NCHRP)
where D A = dead load due to asphalt; I = 0.33 (applied to Project 12-33. Thanks are due to J. M. Kulicki, D. R. Mertz, T. V.
truck portion of the live load only); L = new proposed live Galambos, C. A. Cornell, F. Moses, D. M. Frangopol, R. Green, and
load (see Fig. 5); and )'L = live load factor, 1.6 and 1.7 are I. Friedland for their comments and discussions. The opinions and con-
considered. clusions expressed or implied in the paper are those of the writer and
Reliability indices are calculated for bridge girders de- are not necessarily those of the sponsoring organizations.
signed according to the proposed LRFO code, using (29). Thanks are also due to former and current research assistants at the
University of Michigan for their help in the derivation of load and re-
Various values of resistance factors are considered. The re- sistance models, in particular E-S. Hwang, Y-K Hong, S. W. Tabsh, S-
sults for moments are shown in Figs. 15-17 for steel girders, C. Ting, A. S. Yamani, H. Nassif and T Alberski.
Downloaded from ascelibrary.org by Auburn University on 05/31/16. Copyright ASCE. For personal use only; all rights reserved.

reinforced-concrete T-beams, and prestressed-concrete gird-

ers, respectively. For shear, the results are shown in Figs. APPENDIX. REFERENCES
18-20. Collins. M. P., and Mitchell, D. (1991). Prestressed concrete structures.
The target reliability index was selected equal to 3.5. Con- Prentice-Hall, Inc., Englewood Cliffs, N.J.
sequently, the proposed resistance factors are as follows: <I> Ellingwood, B., Galambos, T. V., MacGregor, J. G., and Cornell, C.
= 0.95 for steel girders, moment, and shear; <I> = 1.00 for A. (1980). "Development of a probability-based load criterion for
American national standard A58." NBS Spec. Publ. 577, Nat. Bureau
prestressed-concrete girders and moment; <I> = 0.90 for rein- of Standards, Washington, D.C.
forced-concrete T-beams and moment; and <I> = 0.85 for Hwang, E-S., and Nowak, A. S. (1991). "Simulation of dynamic load
reinforced-concrete, prestressed-concrete, and shear. for bridges." 1. Struct. Engrg., ASCE, 117(5), 1413-1434.
The proposed load and resistance factors result in an uni- Kulicki, J. M. et al. (1993). "Development of comprehensive bridge
form safety level through a wide spectrum of spans. specifications and commentary." Final Rep., NCHRP 12-33, Modjeski
& Masters, Harrisburg, Pa.
Madsen, H. 0., Krenk, S., and Lind, N. C. (1986). Methods ofstructural
CONCLUSIONS safety. Prentice-Hall, Inc., Englewood Cliffs, N.J.
A probability-based procedure was applied to calculate load Melchers, R. E. (1987). Structural reliability analysis and prediction. Ellis
Horwood Ltd., Chichester, England.
and resistance factors for a bridge design code. The statistical Nowak, A. S. (1993). "Live load model for highway bridges." J. Struct.
basis for load and resistance models was summarized. The Safety, 13(1 +2),53-66.
acceptance criterion for design provisions is closeness to the Nowak, A. S., and Hong, Y-K. (1991). "Bridge live load models." 1.
target safety level. Reliability is measured in terms of the Struct. Engrg., ASCE, 117(9),2757-2767.
reliability index calculated using an iterative procedure. Nowak, A. S., and Lind, N. C. (1979). "Practical bridge code calibra-
Reliability indices were calculated for girder bridges de- tion." 1. Struct. Div., ASCE, 105(12),2497-2510.
Nowak, A. S., Yamani, A. S., and Tabsh, S. W. (1993). "Probabilistic
signed according to the current AASHTO code (Standard models for resistance of concrete bridge girders." AC1 Struct. J., 91(3),
1992). Values of 13 show a considerable variation depending 269-276.
on material, span length, and girder spacing. The calculations Ravindra, M. K., and Galambos, T V. (1978). "Load and resistance
are performed for steel girders, reinforced-concrete T-beams, factor design for steel." J. Struct. Div., 104(9), 1337-1353.
and prestressed-concrete AASHTO-type girders. Standard specifications for highway bridges. (1992). Am. Association of
The proposed LRFO code (Kulicki et al. 1993) includes State Hwy. and Transp. Officials (AASHTO), Washington, D.C.
Tabsh, S. W., and Nowak, A. S. (1991). "Reliability of highway girder
new models for live load, dynamic load, and resistance. The bridges." 1. Struct. Engrg., ASCE, 117(8),2373-2388.
design live load provides a uniform ratio of mean to nominal Thoft-Christensen, P., and Baker, M. J. (1982). Structural reliability
(bias factor). The dynamic load is specified as a constant theory and its applications, Springer-Verlag Inc., New York, N. Y.
fraction of the design truck effect (no dynamic load factor is Ting, S-C., and Nowak, A. S. (1991). "Effect of rebar area loss on
applied to the uniformly distributed portion of the live load). flexural behavior of RIC beams." ACI Struct. J., 88(3), 309-314.
Major changes are introduced in girder distribution factors Ting, S-C., and Nowak, A. S. (1991). "Effect of tendon area loss on
flexural behavior of PIC beams." J. Struct. Engrg., ASCE, 117(4),
(GOFs). New GOFs show a good agreement with analytical 1127-1143.
and experimental results. Resistance is calculated using the Zokaie, T, Osterkamp, T. A., and Imbsen, R. A. (1991). "Distribution
state-of-the-art methodology, e.g., shear design follows the of wheel loads on highway bridges." Final Rep. NCHRP 12-26(1/),
modified compression field theory (CoIlins and Mitchell 1991). Transp. Res. Board, Washington, D.C.

JOURNAL OF STRUCTURAL ENGINEERING I AUGUST 1995 I 1251

J. Struct. Eng., 1995, 121(8): 1245-1251