Está en la página 1de 16

STRUCTURAL

SAFETY

Structural Safety 27 (2005) 230245


www.elsevier.com/locate/strusafe

Deterministic and reliability analysis of prestressed


concrete bridge girders: comparison of the Chinese,
Hong Kong and AASHTO LRFD Codes
J.S. Du a, F.T.K. Au

b,*

School of Civil Engineering, Beijing Jiaotong University, Beijing 100044, China


Department of Civil Engineering, The University of Hong Kong, Pokfulam Road, Hong Kong, China
Accepted 29 October 2004

Abstract
This paper compares the prestressing requirements and reliability indices of prestressed concrete bridge
girders designed using three codes: the Chinese Code, the Hong Kong Code and the AASHTO LRFD
Code. Typical post-tensioned concrete girders of spans ranging from 25 to 40 m are considered. Deterministic analysis indicates that the service limit state governs the design according to the Chinese Code and the
AASHTO LRFD Code. However for the Hong Kong Code, only those with longer spans are controlled by
the service limit state. The actual number of strands needed by the AASHTO LRFD Code is quite close to
that needed by the Chinese Code, while that required by the Hong Kong Code is about 1833% higher than
that required by the AASHTO LRFD Code. Disparity between reliability indices for exural capacity
based on the requirements of the service and strength limit states exists in all three codes. However, the
disparity does not follow the same trends as that of the required number of strands for service and strength
limit states in the three codes. In addition, the reliability index for exural capacity according to the requirements of the service limit state is always higher than that of the strength limit state in the three codes. The
actual reliability indices for exural capacity of the girders considered according to the three codes, which
are governed by the service limit state, are close to one another.
 2004 Elsevier Ltd. All rights reserved.
Keywords: Deterministic analysis; Limit state design; Prestressed concrete structures; Reliability analysis
*

Corresponding author. Tel.: +852 2859 2650; fax: +852 2559 5337.
E-mail address: francis.au@hku.hk (F.T.K. Au).

0167-4730/$ - see front matter  2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.strusafe.2004.10.004

J.S. Du, F.T.K. Au / Structural Safety 27 (2005) 230245

231

1. Introduction
Prestressed concrete bridges have comparatively higher durability and require less maintenance
than other types of bridges. The US national bridge inventory data show that less than 5% of all
existing prestressed concrete bridges are classied as structurally decient, and nowadays prestressed concrete bridges constitute about 50% of the 5000 new projects [1]. In Hong Kong alone,
prestressed concrete bridges amount to about 53% of the total stock [2]. Mainland China is now
undertaking one of the largest infrastructure construction programs in the world, and a lot of new
bridges are built every year. Although no statistical data about the specic percentage of prestressed concrete bridges in Mainland China are available, it is estimated that prestressed concrete
highway bridges also constitute a relatively large part of the new bridge projects. The return of
Hong Kong to China in 1997 has encouraged more communication between them. Several highway projects connecting Hong Kong and Shen Zhen in southern China are now under planning
and being designed. However, bridges in Hong Kong are designed according to the Structures Design Manual for Highways and Railways (SDMHR) [3]. On the other hand, the design of highway
bridges in Mainland China is governed by the Chinese Code [4]. Therefore, there is a need to analyze and compare prestressed concrete bridges designed by these two codes as well as the AASHTO LRFD Code [5] that is commonly accepted in many parts of the world.
The SDMHR Code [3] sets out standards and provides guidance for the design of highway and
railway structures in Hong Kong. The manual is largely based on the British Standard BS5400
[69] and the Departmental Standard BD 37/01 [10] of the United Kingdom Highways Agency,
with some of their provisions modied to suit Hong Kong conditions. Both the Chinese and Hong
Kong Codes are limit state design codes with philosophies broadly comparable to the corresponding British Standards [69], and the two limit states adopted are the strength (ultimate) limit state
and the service (serviceability) limit state.
Prestressed concrete structures are typically designed to ensure that the initial and nal stresses at service load conditions are within the allowable limits, while the ultimate exural capacity of section is checked later. Usually the service limit state governs the design of prestressed
concrete bridge girders. Therefore, the reliability indices of prestressed concrete girders at the
strength limit state for exural capacity depend on the actual number of prestressing strands,
which fulls the requirements of the service limit state. Nowak et al. [11] compared the reliability levels of prestressed concrete girders designed using three dierent codes, and concluded
that the AASHTO LRFD Code [5] provided the most uniform reliability level. In their study,
the exural capacity of prestressed concrete girders was determined by factored load eects,
but the service stress criteria were not considered. Therefore, the actual dierence in reliability
of prestressed concrete bridge girders designed by dierent codes is not yet very clear. As the
AASHTO LRFD Code [5] is a reliability-based code, this paper also adopts it as a reference
for comparison.
The objective of this paper is therefore to compare the reliability indices of prestressed concrete
bridge girders designed using three codes: the Chinese Code [4], the Hong Kong SDMHR Code
[3] and the AASHTO LRFD Code [5] through a comparison of the required number of prestressing strands. Typical post-tensioned prestressed concrete T-girders used in China with dierent
spans are considered. Both the service limit state and the strength limit state are taken into account in the study.

232

J.S. Du, F.T.K. Au / Structural Safety 27 (2005) 230245

2. Deterministic analysis
Four simply supported prestressed concrete bridge decks are considered in the present study.
They were designed with typical post-tensioned prestressed concrete T-girders commonly used
in China, as shown in Fig. 1. The span varies from 25 to 40 m while the girder spacing is kept
at 1.6 m. Fig. 2 shows a typical section of the bridge deck studied. In all the cases studied, the
bridge deck has two 3.5 m wide trac lanes, and it is provided with two normal size parapets
on each side. Apart from the T-girders, the bridge deck comprises an 8 cm thick cast-in-place concrete topping with a 4 cm thick asphalt overlay. The girders have nominal concrete cube strength
fcu of 50 MPa while that of the cast-in-place concrete deck is 40 MPa. The concrete cylinder
strength fc0 is assumed to be 0.8fcu. Prestressing is provided by 12.7 mm diameter low relaxation
strands with ultimate strength fpu of 1860 MPa. Intermediate and end diaphragms of thickness 15
cm and center-to-center spacing of 468 cm are provided to connect the girders transversely to assist in load distribution among girders. The half-diaphragms are precast together with the girders.
The connection between adjacent half-diaphragms is achieved by welding of the steel plates already cast into the half-diaphragms. The depths of the diaphragms are 130, 160, 176 and 206
cm for span lengths of 25, 30, 35 and 40 m, respectively. For comparison, the four bridge decks
are considered under the three design codes mentioned above. The purpose of this deterministic
8cm

158cm

h1

Parameters of T-Girder
Span

Parameters (cm)
h2

h3

25m 145 10

98

19

30m 175 10

128

19

35m 200 12

142

28

40m 230 12

172

28

10cm

h2

16cm

h1

h3

36cm

Fig. 1. Dimensions of typical T-girder sections studied.

0.5m

0.5m

7.0m
Asphalt Overlay

Cast-in-place concrete

1.6m
T-Girder

1.6m

1.6m

1.6m

Diaphragm

Fig. 2. Cross-section of bridge deck studied.

J.S. Du, F.T.K. Au / Structural Safety 27 (2005) 230245

233

analysis is to compare the number of prestressing strands required by the exural moments at the
service and strength limit states as specied by the three codes considered.
2.1. Unfactored bending moments due to dead load and live load
Highway bridges are subjected to dierent types of loads during their design lives, which include dead load, live load (including impact), environmental loads (temperature, wind, earthquake), and other loads (collision, braking, etc). For superstructures of short to medium spans,
environmental loads do not normally govern. Therefore in this study, only dead load and live load
are considered. The dead load mainly comprises the self-weight of precast girders (D1), the weight
of the cast-in-place concrete including the parapets (D2), and the asphalt overlay (D3). Using the
unit weights specied in the respective design codes and assuming that the girders evenly share the
dead load, the unfactored bending moment per girder due to dead load is calculated and the results are listed in Table 1.
There are considerable dierences in the values of design live load specied by the three codes.
In the Chinese Code [4], all bridges are primarily designed for truck loading and then checked for
trailer loading. Four kinds of truck loading as well as the corresponding trailer loading are used in
bridge designs according to the class of highway. For expressways, the design truck loading
Super-20 is in the form of a train of trucks comprising a single 550 kN heavy truck and several
200 kN standard trucks at specied intervals, which has an arrangement of axles as shown in
Fig. 3. For concrete bridges with span length L (m), the dynamic load allowance (impact allowance) I is specied as 0.3 of the static live load for L 6 5 m, 0.0075 (45-L) of the static live load
Table 1
Unfactored dead load moment per girder from deterministic analysis (kN/m)
AASHTO LRFD

MD1

MD2

MD3

MD1

MD2

MD3

MD1

MD2

MD3

976
1520
2452
3468

218
383
522
676

93
164
224
290

976
1520
2452
3468

209
367
500
648

91
160
219
284

937
1459
2354
3329

218
383
522
676

91
160
219
284

10m

70kN
130kN

70kN
130kN
4m

7m

15m

4m

140kN
140kN

10m

1.4m

70kN
130kN

15m

30kN
120kN
120kN

15m

3m
1.4m

Heavy Truck
550kN

Standard Truck
200kN

4m

m
m
m
m

Hong Kong Code

70kN
130kN

25
30
35
40

Chinese Code

4m

Bridge span

15m

Fig. 3. Design live load specied by Chinese Code [4] (not to scale).

234

J.S. Du, F.T.K. Au / Structural Safety 27 (2005) 230245

for 5 m 6 L 6 45 m, and taken as zero for L P 45 m. The trailer loading is only considered when
it results in load eects more severe than those caused by the truck loading. In the cases examined
in this paper, the trailing loading is not critical and is therefore ignored.
The design live load according to the Hong Kong SDMHR Code [3] is either type HA loading,
or type HA loading combined with type HB loading, whichever is more severe in its eects. Type
HA loading is a formula loading representing normal trac. Both types of loading have included
the static component and the dynamic load allowance (impact allowance). The carriageway is rst
divided into a number of notional lanes solely for the purpose of applying the specied live loads.
For the global design of a bridge, type HA loading per notional lane comprises a Knife Edge Load
(KEL) of 120 kN and a Uniformly Distributed Load (UDL) of intensity w kN/m that is dependent on the loaded length L (m), as shown in Fig. 4. The UDL in kN/m per notional lane is derived from the following equation:
8
0:67
>
< w 400  1=L ; L 6 70 m;
1
w 44  1=L0:15 ; 70 < L 6 1400 m;
>
:
w 14:85;
L > 1400 m:
In BD 37/01 [10], the UDL is 336(1/L)0.67 kN/m for loaded lengths up to and including 50 m, and
36(1/L)0.1 kN/m for loaded lengths between 50 and 1600 m. For the loaded lengths investigated in
the present study, which do not exceed 40 m, the UDL in the SDMHR Code is about 19% higher
than that of BD 37/01. A nominal HB loading is a unit loading representing a single abnormal
heavy vehicle having four heavy axles, each of which comprises four wheels, with the spacing between the two interior axles variable in steps. One unit of type HB loading is 10 kN per axle, as
shown in Fig. 5. Generally 45 units of type HB loading shall be used, but for the service limit state,
25 units of type HB loading may be used when calculating crack widths in reinforced concrete
bridges and investigating the exural tensile stresses in prestressed concrete bridges.
The design vehicular live load HL-93 specied by the AASHTO LRFD Code [5] is a combination of a three-axle design truck and a uniform design lane load of 9.3 kN/m as shown in Fig. 6.
The dynamic load allowance IM is specied as 0.33 of the truck load only, with no dynamic load
allowance applied to the lane load. In addition, the code also species the girder distribution

Knife Edge Load (KEL)


120kN per notional lane

Uniformly Distributed Load


(UDL) w kN/m per notional lane

Fig. 4. Design HA live load specied by Hong Kong Code [3].

10kN 10kN
10kN 10kN
1.8m
1.8m 6 to 26m

Fig. 5. Design unit HB live load specied by Hong Kong Code [3] (not to scale).

J.S. Du, F.T.K. Au / Structural Safety 27 (2005) 230245

235

145kN
145kN
35kN
4.3m
4.3 to 9.0m
9.3kN/m

Fig. 6. Design vehicular live load HL-93 specied by AASHTO LRFD Code [5] (not to scale).

factor GDF for use in structural analysis. For evaluation of the bending moments of interior girders with two or more design lanes loaded, the GDF is a function of girder spacing, span length and
stiness of the girder as follows:
!0:6 !0:2
!0:1
S
S
Kg
;
2
GDF 0:075
2900
L
Lt3s
where S is the girder spacing (mm), L is the span length (mm), ts is the thickness of slab (mm), and
Kg is the longitudinal stiness parameter (mm4). The code also recommends modied girder distribution factors for exterior girders.
The maximum unfactored live load moments per girder were calculated according to the design
live load specied by the three codes, and the results including dynamic load allowance are listed
in Table 2. In the determination of the maximum live load moment per girder for live load specied by the Chinese code and Hong Kong Code, the GuyonMassonnet method [12] for load distribution was used. For calculations with the Hong Kong Code, type HA loading with 45 units of
type HB loading governs the design of strength limit state, while type HA loading with 25 units of
type HB loading controls the design of service limit state. Under the loading specied by the Chinese code and Hong Kong Code, the exterior girders are more heavily loaded. The AASHTO
LRFD Code actually provides girder distribution factors for calculation of live load moment
for individual girders. For the cases examined, the maximum live load moment per girder is
the same as those of the interior girders. Apart from calculating the maximum live load moments
according to the code, they were also calculated using the GuyonMassonnet method for comparison as for the other two codes. It is noticed that the dierences between the two sets of results
are rather small, as shown in the last column of Table 2. In the subsequent analysis, the results
Table 2
Maximum unfactored live load moment per girder from deterministic analysis (kN/m)
Bridge span

MLL+IM
Chinese Code

25
30
35
40

m
m
m
m

1254
1619
2084
2496

Hong Kong Code

AASHTO LRFD Code

HA + HB45

HA + HB25

2932
3788
4603
5454

1668
2164
2635
3125

1564
2037
2476
2986

(1500)
(1957)
(2425)
(2925)

Note: Live load moments calculated from loading specied by the AASHTO LRFD Code using the GuyonMassonnet
method are shown within parentheses.

236

J.S. Du, F.T.K. Au / Structural Safety 27 (2005) 230245

calculated from the girder distribution factors given by the code were adopted for load combination and further analysis.
2.2. Bending moments for service and strength limit states
The specic load factors (partial load factors) for the service and strength limit states in the
three codes are dierent. Even in the same code and for the same limit state, the load factors
may also be dierent for dierent design live load. For example, in the Hong Kong Code, the
load factors for type HA loading are dierent from those for type HB loading. This paper
only lists the load factors or formulas, which are relevant to the cases considered. The bending
moments at the service limit state are mainly used to check the tensile stresses in concrete,
while those at the strength limit state mainly govern the minimum exural resistance required.
The design bending moments Ms and Mu for the service and strength limit states, respectively,
are as follows:
(a) Chinese Code
M s M D1 M D2 M D3 M LLIM ;

3a

M u 1:251:2M D1 M D2 M D3 1:4M LLIM :

3b

(b) Hong Kong Code


M s M D1 M D2 1:2M D3 1:1M LLIMHAHB25 ;

4a

M u 1:11:15M D1 1:5M D2 1:75M D3 1:3M LLIMHAHB45 :

4b

(c) AASHTO LRFD Code


M s M D1 M D2 M D3 0:8M LLIM ;

5a

M u 1:25M D1 M D2 1:5M D3 1:75M LLIM ;

5b

where MD1, MD2 and MD3 are the dead load moments per girder for groups D1, D2 and D3,
respectively, and MLL + IM is the live load moment per girder. The values of Ms and Mu required
by the three codes were calculated and shown in Fig. 7. It can be seen that the requirements of the
Chinese Code are fairly close to those of the AASHTO LRFD Code in respect of Ms and Mu.
However, the requirements of the Hong Kong Code are all larger than the corresponding values
of the other two codes.
2.3. Comparison of prestressing requirements
The prestressing requirements of dierent codes are most conveniently compared by examining
the required number of prestressing strands and this number must be an integer. Both the Chinese
and Hong Kong Codes require that no tensile stresses should appear in the precompressed tensile
zone of prestressed concrete components for the combination involving mainly dead load and live

J.S. Du, F.T.K. Au / Structural Safety 27 (2005) 230245

237

Fig. 7. Bending moment per girder for strength and service limit states.

load. In other words, the allowable tensile stress in concrete is p


zero.
0 According to the AASHTO
LRFD Code, the allowable concrete tensile stress in MPa is 0:5 f c for components with bonded
prestressing tendons
p0 or reinforcement that are subjected to not worse than moderate corrosion
conditions, 0:25 f c for similar cases but with severe corrosive conditions, and zero for components with unbonded prestressing tendons, where fc0 is the concrete cylinder strength. For ease
of comparison in this study, the allowable concrete tensile stress is taken as zero for all three
codes.
For ease of comparison, the ultimate bending moment Mu is calculated taking into account the
prestressing tendons only but ignoring the non-prestressed reinforcement. The specied tensile
strength fpu of prestressing steel is taken as 1860 MPa in this paper while the cross-sectional area
of prestressing steel Aps is to be worked out. The initial stress in prestressing steel fpi is taken to be
0.70 fpu or 1302 MPa. Assuming a loss of prestress of 20% of the initial prestress, the eective
stress in prestressing steel fpe is therefore 1042 MPa. For the concrete with cube strength fcu of
50 MPa, the concrete cylinder strength fc0 is taken as 0.8 fcu or 40 MPa. Although the basic approach to determine the number of prestressing strands to satisfy the requirements of the strength
limit state for bending is the same, the three codes considered all have dierent design formulas to
cope with dierent design material strength and material factors. For beams with rectangular sections or anged sections with a breadth or width of compression ange b in which the depth of
stress block falls within the thickness of the compression ange, the required number of prestressing strands can be calculated from the following equations of equilibrium of forces and moments,
respectively, which have already incorporated the necessary factors:

238

J.S. Du, F.T.K. Au / Structural Safety 27 (2005) 230245

(a) Chinese Code


fcu0 bx Aps fpy ;

6a

M u Aps fpy d  0:5x;

6b

where fcu0 is the average compressive stress of the equivalent rectangular stress block, x is the depth
of equivalent rectangular stress block, fpy is the design yield strength of prestressing steel, and D is
the distance from the extreme compression ber to the centroid of prestressing steel. For the concrete with cube strength fcu of 50 MPa, from rst principles or a table in the Chinese Code, fcu0 is
determined as 28.5 MPa. The design yield strength fpy of prestressing steel is taken as 0.9 fpu or
1674 MPa for low relaxation strands.
(b) Hong Kong Code
0:4f cu bx Aps fpb

7a

M u Aps fpb d  0:5x;

7b

where fpb is the tensile stress in the tendons at exural failure of the section, which can be obtained
from rst principles or a table in the code BS5400:Part 4 [8].
(c) AASHTO LRFD Code
0:85b1 fc0 b kAps fpu =d p c Aps fpu ;

8a

M u Aps fps d p  0:5b1 c;

8b

where fc0 is the concrete cylinder strength taken as 40 MPa, c is the distance between the neutral
axis and the compressive face or the neutral axis depth, and the yield strength of prestressing steel
fpy is similarly taken as 0.9 fpu or 1674 MPa for low relaxation strands. Also relevant to the above
two equations are the stress block factor b1, the average stress in prestressing steel fps and the factor k given, respectively, as
0:85 P b1 0:85  0:05fc0  28=7 P 0:65;

8c

fps fpu 1  kc=d p ;

8d

k 21:04  fpy =fpu :

8e

In the present case, these factors are worked out as b1 = 0.76 and k = 0.28.
Analyses for the service and strength limit states are carried out for the above bridge girders
with span lengths ranging from 25 to 40 m according to the above three codes. They include
not only the evaluation of the required number of prestressing strands for the service
and strength limit states, but also the checks to ensure that the concrete stresses are within

J.S. Du, F.T.K. Au / Structural Safety 27 (2005) 230245

239

allowable limits. As shown in Fig. 8, the number of strands required to satisfy the service limit
state is always greater than that needed for the strength limit state in the Chinese Code and the
AASHTO LRFD Code. For design in accordance with the Hong Kong Code, the number of
strands required to satisfy the service limit state is above that needed for the strength limit
state for span lengths of 35 and 40 m only. It may therefore be concluded that the service limit
state in respect of the allowable concrete tensile stress governs the design according to the Chinese Code and the AASHTO LRFD Code. However for bridge designs to the Hong Kong
Code, those with shorter spans are governed by the strength limit state, while those with longer
spans are controlled by the service limit state. For the same bridge girder, the actual number of
strands required by the Chinese Code is slightly larger than that needed by the AASHTO
LRFD Code by up to 4%. In comparison, the number of strands required by the Hong Kong
Code exceeds that needed by the AASHTO LRFD Code by a large margin, ranging from 33%
to 18% for the span range from 25 to 40 m, respectively.
For convenience in comparison, the term strand ratio g [13] is also adopted here. The strand
ratio g is the ratio of the number of strands required to satisfy the allowable stress requirement
at the service limit state Nser to that needed for the strength limit state Nstr, i.e.
g N ser =N str :

If g is greater than 1, it represents the additional factor of safety the prestressed concrete bridge
girders possess beyond the minimum value required by the strength limit state. Fig. 9 shows the
variation of the strand ratio g with the span length. The variations of the strand ratio g are rather
dierent, being 1.141.17 for the Chinese Code, 0.941.03 for the Hong Kong Code and 1.211.34
for the AASHTO LRFD Code. It might also be inferred that, everything else being the same, with
the increase in initial prestress fpi, say up to 0.75 fpu or 0.80 fpu, the value of g would be decreased

65

Number of Strands

55

45

Chinese Code-Service
Chinese Code-Strength
AASHTO LRFD-Service
AASHTO LRFD-Strength
Hong Kong Code-Service
Hong Kong Code-Strength

35

25

15
20

25

30

35
Span length (m)

40

45

Fig. 8. Number of 12.7 mm strands per girder required for strength and service limit states.

240

J.S. Du, F.T.K. Au / Structural Safety 27 (2005) 230245


1.5

Strand Ratio

1.4

Chinese Code
AASHTO LRFD
Hong Kong Code

1.3
1.2
1.1
1.0
0.9
20

25

30

35

40

45

Span Length (m)

Fig. 9. Strand ratio g in the three codes.

correspondingly, whereas the decrease in initial prestress fpi would lead to the increase of the value
of g.

3. Reliability analysis
To compute the reliability index, it is necessary to obtain rst the statistical parameters including the mean values or bias factors, coecients of variation and probability distribution types of
load and resistance. The bias factor is dened as the ratio of the mean value of a variable to its
nominal or design value.
3.1. Statistical parameters of load and resistance
To be consistent with the above deterministic analysis, it is convenient to classify the dead load
into three components according to their quality control, namely the self-weight of precast girders
(D1), the weight of the cast-in-place concrete including the parapets (D2), and the asphalt overlay
(D3). All components of dead load are treated as normal distribution random variable. For
bridges designed to the AASHTO LRFD Code, the bias factors for groups D1, D2 and D3 are
taken to be 1.03, 1.05 and 1.0, respectively, while their coecients of variation are taken to be
0.08, 0.10 and 0.25, respectively, as concluded by Nowak [14].
The eects of live load depend on many parameters including mainly the span length, truck
weights, axle loading and conguration, number of vehicles on the bridges, transverse and longitudinal positions of the vehicles on the bridge, etc. The statistical model for live load developed by
Nowak and Hong [15], and Nowak [16] is used here. Clarication [17] regarding the previous
study on Eurocode, Spanish Norma IAP and the AASHTO LRFD Code [11] has revealed that
the bias factor of live load moment per girder is about 0.9541.023, while the coecient of

J.S. Du, F.T.K. Au / Structural Safety 27 (2005) 230245

241

variation is 0.12 [18]. Therefore, for the present study of live load moment per girder designed to
the AASHTO LRFD Code, normal distribution is assumed with the bias factor taken as 1.0 and
the coecient of variation taken as 0.12 for a design life of 75 years. Although the design life
of the bridge based on the Chinese Code and the Hong Kong Code are, respectively, 100 and
120 years, the same design life of 75 years is used in the reliability analysis for all bridges for
the sake of consistency.
The statistical parameters of the exural strength or resistance of prestressed concrete bridge
girders have been studied by Tabsh and Nowak [19] for the AASHTO LRFD Code. It is
found to be distributed in a lognormal manner with a bias factor kR of 1.05 and a coecient
of variation VR of 0.075. Table 3 summarizes the types of probability distribution and statistical parameters of the load and resistance for reliability analysis. In general, once a bridge has
been designed by whatever code selected, it is a physical entity and the reliability analysis must
be performed with respect to that physical entity. In order to provide a common basis for the
comparison of reliability indices, the resistance values for designs from dierent codes were always calculated according to the AASHTO LRFD Code based on the number of prestressing
strands obtained from deterministic analysis. In the reliability analysis, all dead load moments
are those calculated using the unit weights of the AASHTO LRFD Code. This means that for
the same bridge, the dead load moments for reliability analysis are simply those for the AASHTO LRFD Code.
3.2. Procedure for reliability analysis
The conventional procedure for rst-order reliability analysis [18] is then followed. The reliability of a bridge girder cross-section for exural capacity at the strength limit state can be expressed
using a performance function g(. . .) dened as
gR; D; L R  D  L;

10

where R is the resistance or load-carrying capacity, D is the dead load eect, and L is the live load
eect including impact. In the study, failure refers to the case when the exural capacity is
reached. Failure occurs if g is less than or equal to zero. Structural safety can be conveniently
measured in terms of the reliability index b dened as
11

b lg =rg

in which lg and rg denote the mean and standard deviation of the performance function g, respectively. The reliability indices of exural capacity are calculated based on girder sections with the

Table 3
Assumed statistical parameters for load and resistance in reliability analysis
Dead load
Distribution type
Bias factor
Coecient of variation

D1

D2

D3

1.03
0.08

Normal
1.05
0.10

1.0
0.25

Live load

Resistance

Normal
1.0
0.12

Lognormal
1.05
0.075

242

J.S. Du, F.T.K. Au / Structural Safety 27 (2005) 230245

amount of prestressing strands required to satisfy the strength and service limit states. In this
study, the reliability index is computed using the RackwitzFiessler procedure [20]. The basic idea
behind this procedure begins with the calculation of the equivalent normal values of the mean
and standard deviation for each non-normal random variable.
3.3. Reliability indices
With the above-mentioned assumptions and statistical parameters and using the procedure for
reliability analysis, reliability indices of exural capacity of sections with prestressing strands
based on the requirements of the strength and service limit states are calculated for the girders
with span lengths ranging from 25 to 40 m according to the three codes. They are shown in Figs.
10 and 11, respectively. The reliability indices based on the requirements of the strength limit state
are 3.94.4 for girders designed by the AASHTO LRFD Code, 5.25.3 for girders designed by the
Chinese Code, and 3.43.5 for girders designed by the Hong Kong Code. The reliability indices in
the Chinese Code are therefore about 2033% higher than those in the AASHTO LRFD Code
and about 52% higher than those in the Hong Kong Code.
The previous deterministic analyses indicate that, for almost all the cases, the service limit state
governs the design of prestressed concrete bridge girders according to the three codes studied. The
only exceptions are the girders of shorter spans designed to the Hong Kong Code, which are only
marginally governed by the strength limit state. Therefore, the reliability index for exural capacity of prestressed concrete bridge girders is practically governed by the requirements of service
limit state. Fig. 11 compares the reliability indices based on the requirements of service limit state
in the three codes. The reliability indices are 6.36.5 for girders designed by the AASHTO LRFD
Code, 6.66.7 for girders designed by the Chinese Code, and 6.46.9 for girders designed by the
Hong Kong Code. As deterministic analyses indicate that the design is practically governed by
the service limit state, the actual reliability indices of the girders designed to the three codes are

Reliability Index

Chinese Code
AASHTO LRFD
Hong Kong Code

6
5
4
3
2
20

25

30

35

40

45

Span Length (m)

Fig. 10. Reliability index for exural capacity based on the requirements of the strength limit state in the three codes.

J.S. Du, F.T.K. Au / Structural Safety 27 (2005) 230245

243

Reliability Index

7
6
5

Hong Kong Code


Chinese Code
AASHTO LRFD

4
3
2
20

25

30
35
Span Length (m)

40

45

Fig. 11. Reliability index for exural capacity based on the requirements of the service limit state in the three codes.

therefore those based on the service limit state. The reliability indices are also fairly close to one
another, as opposed to the marked dierence in the requirements for prestressing strands.
Comparing Figs. 10 and 11, it can be seen that dierences between reliability indices of exural
capacity of sections with prestressing strands based on the requirements of strength and service limit
states exist in each design code, but these dierences do not follow the same trends as the strand
ratio g in the deterministic analysis. Although the variation of the strand ratio g in the Hong Kong
Code (i.e. 0.941.03) is the smallest among the three codes, the Hong Kong Code has the largest
disparity of reliability indices required to satisfy the service and strength limit states among the three
codes studied. In particular, the reliability index based on the requirements of the service limit state
is 83103% higher than that for the strength limit state. For girders designed by the AASHTO
LRFD Code, the reliability index based on service limit state requirements is about 4365% greater
than that for the strength limit state. In the case of the Chinese Code, the reliability index according
to the service limit state is about 2330% higher than that of the strength limit state.
3.4. Discussions
The above reliability analysis indicates that in the design of prestressed concrete bridge girders,
disparity of reliability indices for exural capacity based on the requirements of the service and
strength limit states does exist in all three codes considered. To achieve a uniform safety level
in the design of prestressed concrete bridge girders, reliability-based code calibration at the
strength limit state should also take into account the inuence of service limit state. The use of
radically dierent live loads in the Hong Kong Code, namely the rather heavy HA and 45 units
of HB loading for the strength limit state, and the much lower HA and 25 units of HB loading for
the service limit state, may have contributed to the large disparity of reliability indices. In addition, the high live load moment for the strength limit state in the Hong Kong Code also causes
the reliability index for exural capacity at the strength limit state to be lower than that of the
AASHTO LRFD and the Chinese Codes.

244

J.S. Du, F.T.K. Au / Structural Safety 27 (2005) 230245

4. Conclusions
Deterministic and reliability analyses are performed for simply supported post-tensioned T-girders designed according to three codes, namely the Chinese Code, the Hong Kong Code, and the
AASHTO LRFD Code. They cover spans ranging from 25 to 40 m. Subject to the limitations of
the assumptions and parameters adopted in the study, the following conclusions can be obtained:
1. The service limit state governs the design of prestressed concrete bridge girders in the Chinese
Code and the AASHTO LRFD Code. For the Hong Kong Code, bridges of shorter span
lengths (i.e. 2530 m) are governed by the strength limit state whereas those with longer span
lengths (i.e. 35 m40 m) are governed by the service limit state.
2. The actual number of strands required by the Chinese Code is slightly more than that needed
by the AASHTO LRFD Code. The number of strands required by the Hong Kong Code
exceeds that needed by the AASHTO LRFD Code by a large margin, ranging from 33% to
18% for the span range of 2540 m, respectively.
3. The variations of the strand ratio g for the span range of 2540 m are: 1.141.17 for the Chinese Code, 1.211.34 for the AASHTO LRFD Code, and 0.941.03 for the Hong Kong Code.
4. Reliability analysis indicates that disparity between reliability indices for exural capacity
based on the requirements of the service and strength limit states exists in all three codes. However, the disparity does not follow the same trends as those of the strand ratio g obtained from
the deterministic analysis for the three codes. For girders of span range of 2540 m designed by
the Hong Kong Code, the reliability index for exural capacity based on the requirements
of the service limit state is 83103% higher than that for the strength limit state. For similar
girders designed by the AASHTO LRFD Code, the reliability index based on the service limit
state is about 4365% greater than that for the strength limit state. In the case of the Chinese
Code, the reliability index based on the service limit state is about 2330% above that for the
strength limit state.
5. To achieve a uniform safety level in the design of prestressed concrete bridge girders, reliability-based code calibration should take into account not only the strength limit state but also
the service limit state.
6. The actual reliability indices for exural capacity of the girders considered by the three codes,
which are governed by the service limit state, are close to one another in spite of the large differences in the number of prestressing strands needed.

Acknowledgments
The authors are grateful to Professor A.S. Nowak, Department of Civil and Environmental
Engineering, University of Michigan, USA and Dr. C.H. Park, Research Institute of Industrial
Science & Technology, Korea for useful discussions regarding their work described in [11]. The
authors are also grateful to the referees for their useful suggestions. The work described in this
paper has been partly supported by the Research Grants Council of the Hong Kong Special
Administrative Region, China (RGC Project No. HKU 7101/04E).

J.S. Du, F.T.K. Au / Structural Safety 27 (2005) 230245

245

References
[1] Dunker KF, Rabbat BG. Performance of prestressed concrete highway bridges in the United Statesthe rst 40
years. PCI J 1992;37(3):4864.
[2] Wong PC, Wong CY, Darby J. Bridge management in Hong Kong: the selection of appropriate techniques.
Current and future trends in bridge design, construction and maintenance, vol. 2. London: Thomas Telford; 2001.
p. 312.
[3] Structures Design Manual for Highways and Railways. Highways Department, Government of the Hong Kong
Special Administrative Region, 2nd ed., with Amendment No. 1/2002, Hong Kong; 1997.
[4] Chinese Design Code for Highway Bridges. Beijing: Peoples Communication Press; 1991.
[5] AASHTO LRFD Bridge Design Specications. Washington, DC: Association of State Highway and Transportation Ocials; 1998.
[6] BS5400. Steel, concrete and composite bridges. Part 1: General statement. London: British Standards Institute;
1978.
[7] BS5400. Steel, concrete and composite bridges. Part 3: Code of practice for the design of steel bridges. London:
British Standards Institute; 1982.
[8] BS5400. Steel, concrete and composite bridges. Part 4: Code of practice for the design of concrete bridges. London:
British Standards Institute; 1990.
[9] BS5400. Steel, concrete and composite bridges. Part 5: Code of practice for the design of composite bridges.
London: British Standards Institute; 1979.
[10] BD 37/01. Loads for highway bridges. Design manual for roads and bridges. London: the Stationery Oce Ltd.;
2001.
[11] Nowak AS, Park CH, Casas JR. Reliability analysis of prestressed concrete bridge girders: comparison of
Eurocode, Spanish Norma IAP and AASHTO LRFD. Struct Safety 2001;23:33144.
[12] Bares R, Massonnet C. Analysis of beam grids and orthotropic plates by the GuyonMassonnetBares
method. New York: Frederick Ungar Publishing Co.; 1968.
[13] Tabsh SW. Reliability based parametric study of pretensioned AASHTO bridge girders. PCI J 1992;37(5):5667.
[14] Nowak AS. Calibration of LRFD bridge code. ASCE J Struct Eng 1995;121(8):124551.
[15] Nowak AS, Hong YK. Bridge live load models. ASCE J Struct Eng 1991;117(9):275767.
[16] Nowak AS. Live load model for highway bridges. Struct Safety 1993;13(12):5366.
[17] Park CH. Private communication; 2002.
[18] Nowak AS, Collins KR. Reliability of structures. Boston: McGraw-Hill; 2000.
[19] Tabsh SW, Nowak AS. Reliability of highway girder bridges. ASCE J Struct Eng 1991;117(8):237288.
[20] Rackwitz R, Fiessler B. Structural reliability under combined random load sequences. Comput Struct
1978;9(5):48994.

También podría gustarte