192 lvkovic and Perisic

0
3Li0 -2R. JR. dy.1 dw(x, t 0 )
H.(t ) E A - -d- d dx -
10 ii OX X

1 IR. dw(x, to) 2

- 2 ( dx ) dx; i = 1,2. (18)
0

With solution (14), (15) or (16) and conditions (18), the

state of stress and strain for instantaneous elastic solution of the
two-chord system is completely determined.

In the second step, with H1(t 0 ), H2(t 0 ) and w(x,t 0 ) already

known, it is possible to determine H1(t), H2(t) and w(x,t) for any
time t.
In this case equation (12) can be written in the form:
4 2 a
1 d w(x,t) + K2 d w(x,t) -q>(x) + (___!_ _ 1 ) ljl(x,t) = o, (19)
bt dx 4 t dx2 bt o
4
d w(x,t ) H <t>- H (t)
0 2 1 2
where 1jJ (x,to) = dx4 and Kt R1 + R2

The general solution then has the following form, depending

on the value of Kf:
2
a) for Kt > 0

w*(x, t) = C~t sinKtx + c;t cosKtx +

a
+ c3tx a + [ l.p(x)dx] dx-
+ c4t ss-
- ( :~ - 1) J[J~ (x , t 0
) dx] dx (20)

2
b) for Kt < 0

w*(x,t) =C~tShKtx+C~tChKtx +

+ c~tx + c~t +ScJi:p(x)dxJ dx-

-(at - 1) JcJ~ (x )dxJ dx (21)

bt
2
c) for K = 0
t
the function w* ( x, t) may be obtained by direct integration of the
equation:
 Prestressing 193

where (p(x) - particular integral of the differential equation (19),

1
and w*(x,t) = b w(x,t).
t
Now, when w(x,t) is obtained from equations (20), (21) or
(22), horizontal forces may be obtained directly from conditions
(17).

PRESTRESSED TWO-CHORD SYSTEMS WITH PRESTRESSING

TENDONS OUTSIDE THE CONCRETE CROSS SECTION

A prestressed two-chord system with an upper compressed

reinforced concrete chord with considerable bending stiffness,
usually straight, or polygonal upward convex, and a lower tensio-
ned chord upward concave, consisting only of prestressing steel
or a combination of rigid steel members and prestressing steel,
with a negligible bending stiffness, will be analyzed next. Fig. 3
shows such a structural system with a tensioned chord made only
of prestressing tendons.

In such a case, the determination of the state of stress and

strain is considerably simplified compared to the already shown
general solution and is reduced to the solution of differential
equation ( 12), taking into consideration that:

and using conditions ( 17) , where the lower chord does not pos-
sess the property of creep.

Here too, the problem is solved in two steps. First, the

stress-strain state for t=t 0 (instantaneous elastic problem) has to
be solved. With w(x, t 0 ), H1 (t 0 ), and H2(t 0 ) obtained, w(x, t) and
H1(t) and H2(t) are defined by means of slightly modified equa-
tions (14),(15) or (16); that is, (20),(21) or (22),with "length"
conditions (17) and (18) .

For vertical load only, when K~0 =K~, the solution of the
stress strain state is reduced to the integration of equations (16)
and (22), in order to define w(x,t 0 ) and w(x,t} with considerably
simplified "length" conditions ( 17) and (18) .

For the case shown in Fig. 3 :

0 0 - 4f2
y 0, y - x(R.-x).
1 2 - R, 2
194 Ivkovic and Perisic

As the load is only vertical, H1=H2. For the action of dead

load and prestressing force (case t=t ) , the solution of equation
(16) gives: 0

1 H (t ) 2 4 3 3
wg(t 0 )= 24 R [ g - 8 7 f2 J[x -2xR.+xR. J. (23)
1
The length conditions, for t=t 0 , that is, at=bt=1, using
expression ( 17) , are:
3L -2R. 3L -2R.
H (t )[ 1o + 2o J -
g o E1~1 E2A2

- rR, dy ~ ~ dx - \R, (dw g ) 2 dx = 0 . (24)

J0 dx dx J0 dx

In many practical cases for typical shallow catenaries for

f/R-<1/10, it is possible to neglect the third term of equation (24),
compared to the first two terms. The length condition is then:
0
3L 10 -2R- 3L 20 -2 R. JR. dy2 dw
H (t ) [ E A + E A J- -d- ____g__
dx
dx = 0 , (25)
go 11 22 ox
and

(26)

However, if the thir1 ~erm of equation (24) can not be

neglected, the value for Hg2 (t 0 ) is obtained as one of the roots
of the square algebraic equation in the form:
2
H( 2 ) (t ) -a H( 2 ) ( t ) +f3 =0 (27)
g 0 t g 0 t
c c
where a = [(-1- + _2_) R2 + 8 f 2 R R. + 16D R. 5 f g] 1
T KN1 KN2 1 15 2 1 2 64Df2R. 3 '
2
g 1 3 7
f3 1 = 3 [ 15 R1F2R- + Dg R. J
64D2R.
2
and, in this case, D = 8.4325·10-4,
C = 3L -2R., c = 3L -2R..
1 10 2 20

It is easy to show that H~ )(t 0) with Wg(t 0 ) from expression

1

(23) presents the solution to the problem according to the First

Order Theory, when the influence of deformations of the system
upon the equilibrium of forces acting on the girder is neglected.
The solution HJ2) (t 0 ), from equation (27), with wg(t 0 ) from (23),
presents the s8lution of the problem according to tfie Second
 Prestressing 195

Order Theory. The comparison of the H~1 )(t 0 ) and H~2 )(t0 )
values shows whether the solution accoratng to the First Order
Theory Is sufficiently accurate.

When the problem Is solved for t=t 0 , It Is easy to obtain

wg (t) for any t>t 0 and corresponding at and bt, using the equation
(22) and the value Hg(t) from modified conditions (17) for the
Theory of First, as well as for the Theory of Second Order. Then:
1 8f2
wg(t) = 24R1 {bt [ g - 7 Hg(t)] +

+ (at-bt) [g- :~2 4 3 3

Hg(t 0 )J} (x -2x R. +x£ ). (28)

According to the conditions (25), that Is, modified condltlon

(17), with wg(t) from (28) the following expression Is obtained:

1 3 8 2 R1
atg I5 f2£ - (at- bt) Hg(to) [ 15 f2 £ + E A C1 J
H(l) (t) 1 1 (29)
g 8 2 R1 R1
bt [ 15 f2 £ +EA C1J +EA c2
1 1 2 2

The equation (27) and condition (24) give:

H(2) (t) [g - Pt .; ] 8£f2 ' (30)

g t 2
where
f
Pt = bt [ g - 8 ; H (t)]
R, g

EXAMPLE OF APPLICATION: THE NEW PRESTRESSED CONCRETE

HANGAR AT THE BELGRADE INTERNATIONAL AIRPORT
IN YUGOSLAVIA

The new prestressed concrete hangar at the Belgrade Inter-

national Airport was designed for the simultaneous maintenance
of two Boeing 747 jets. The authors of this paper are the main
designers of the hangar structure together with A. Pakvor and
M. Aetc, professors of Civil Engineering at the Department of
Civil Engineering at the University of Belgrade. The hangar was
finished In 16 months, In 1986.

Fig. 4 shows the general plan of the hangar which is of a

rectangular base, 135.80 by 70.05 m. The door of the hangar
provides 21.90 m clear height. It is situated along the entire
longer side of the base, which is without any intermediate columns.
The door has six wings moving along the three ralls, so that two
196 lvkovic and Perisic

thirds of the door may be simultaneously opened. It is foreseen

that Boeing 747 jets enter the hangar with their front parts. In
order to save the necessary space, two 16.80 by 22.40 m recesses
have been designed in the rear hangar wall, increasing the overall
hangar depth to 92. 45 m.

The main structural members of the hangar are the three

prestressed reinforced concrete box girders, of 135.80 m span,
parallel to the longer side of the base, spaced at 22.40 m. The
main girders are laid on reinforced neoprene bearings on the top
of the reinforced concrete main columns in the side walls of the
hangar. The height of the main columns in the portal plane is
36.35 m, while the other two pairs of main columns are lower, for
0. 90 m each. The highest level of the hangar structure is 37. 35 m
and was governed by the flying safety regulations in the Airport
zone.

The complete roof structure of the hangar is hung on the

main girders. In this way the minimum surface of the facade walls
and the minimum effective volume have been achieved, thus redu-
cing the service costs for the winter heating of the hangar. The
roof structure consists of reinforced concrete secondary girders
spaced at 8. 40 m and purlins at 3. 75 m. Light-weight concrete
plates together with thermo- and hydro-insulation and plastic
lanterns are placed over the purlins.

The hangers carrying the roof structure are 50 mm diameter

steel rods. Two pairs of such hangers are placed at 8. 40 m along
the girders span. They transmit the roof weight to the special
supports inside the main box girders. The hangers have threads
with nuts at their ends by which the level of the complete roof
can be easily adjusted from inside the reinforced concrete girder,
when, and 1f necessary.

The hangar door is a steel structure. At its upper edge

the door rests horizontally on the roof structure, over special
steel guides. The wind and seismic forces perpendicular to the
door of the hangar are transmitted to the main columns and rein-
forced concrete walls in the rear wall recesses by means of a
horizontal steel brace in the roof plane. The horizontal forces per-
pendicular to the side hangar walls are accepted by the frame
system, consisting of the main girders and columns, together with
the structure of the rear wall of the hangar.

The foundations of the hangar have been designed for con-

struction directly on the loose soil and are rather shallow. The
foundation level of the main columns is only 3. 25 m under the
hangar floor and the foundations surface is 11. 00x7. 50 m.

The main prestressed reinforced concrete box girders shown

in Fig. 5, are two-chord structural systems, with a span of
135.80 m and a maximum rise of 9. 70 m. The upper compressed
chord is made of the fc' = 45 MPa reinforced concrete. The cross-
 Prestressing 197

section of the box girder is constant along the whole span, having
4. 00 m external width and 2. 80 m external height. The upper slab
is 35 em thick, the side walls are 25 em and the thickness of the
bottom slab is 20 em. The percentage of the mild reinforcement in
the box girder along the whole span is about one percent.

The lower chord is made of 27 tendons consisting of 11

strands 15.2 mm in diameter each. The applied prestressing sys-
tem is the Yugoslav IMS (Institute for Testing of Materials of
Serbia, Belgrade) system which has been successfully used for
over 35 years in Yugoslavia and some other countries.

The admissible force in a strand is 1, 786 kN. Three tendons

of 11 0 15. 2 mm strands are placed together in a single polyethilene
protective tube, 142 mm in diameter, so that there are 9 such
composite tendons along the whole span. However, near the ends
of the girders the tendons are split up and separately anchored
using 27 standard permanent anchors.

The designed polygonal tendon configuration is provided by

7 pyramidal "chairs" made of steel pipes, oriented in the direction
of the line of symmetry of intersecting angles of the lower chord.
The tendons are led across the chairs over special rotational
bearings which reduce the friction coefficient to about 0. 04. This
resulted in minimal frictional losses.

The general view of the prestressing tendons and hangers

carrying precast roof elements is shown on Fig. 6, taken during
the hangar construction.

The structural system of the main girders, shown in Fig. 5,

particularly for such large spans, may be understood as a two-
chord catenary system, discussed in the first part of this paper.
The upper reinforced concrete chord has axial and bending stiff-
nes, while the lower steel chord can carry only tension. The
chords are interconnected at suitable distances by compressed
elements. For shorter spans, such a structural system may be
simply understood as a reinforced concrete girder on elastic sup-
ports. The stiffness of supports depends on the tensile capacity
and the configuration of the lower prestressing steel chord.

By prestressing such a structural system, the compressive

force is introduced into the upper reinforced concrete chord, but
at the same time "lifting" forces appear at the points of the sup-
porting chairs.

By designing the external prestressing tendons outside the

basic reinforced concrete cross section of the main girders , a very
rational and economical structural system has been obtained [2].
The efficient corrosion protection in the contemporary external
prestressing, allows for tendons to be left free in space. The
main characteristic of such structural systems is that a relatively
high rise may be given to the tendons, increasing very much
their efficiency. In such two-chord structural systems, the rein-
198 lvkovic and Perisic

forced concrete chord is exposed to relatively high normal com-

pressive forces and small bending moments, so that it is in the
stress state free of cracks. For dead loads such structural systems
are not subjected to tensile stresses at all. Therefore, the most
appropriate task is fulfilled by each of the materials - the concrete
withstands only compressive stresses while the steel is exposed
only to tensile stresses.

Due to the considerable decrease of dimensions, that is, the

area of the cross section, such girders are relatively light-weight.
They are exceptionally favorable for significant dead loads and
relatively small, mostly symmetrical live loads, as it is usually the
case with roof structures of medium and very large spans [3],
but can also be successfully used for highway bridges.

The analysis of the main girders was done according to the

First and Second Order Theories, using a computer program which
provided the possibility of considering the creep, shrinkage and
relaxation effects. In spite of the large span and a considerable
reduction of the stiffness of the reinforced concrete part of the
system, the results obtained by both theories did not differ by
more than eight percent. The analysis was performed using the
algebraic stress-strain relationship for concrete, and the Age-
Adjusted Effective Modulus Method (AAEMM). The influence of
creep and shrinkage of concrete on time-dependent behaviour of
the girders was analysed for the supposed load histories, depend-
ing mostly on the construction method. The minimum and maximum
possible values of the expected creep and shrinkage of concrete
were introduced into the design.

The characteristic cross sections of the main girders were

dimensioned according to the limit states design. The limit state
of deformations was analyzed very carefully, considering that due
to the significant reduction of stiffness, such systems are relatively
deformable. Nevertheless, the calculated midspan deflection caused
by the snow, is only 15 em, due to the low live load to the total
dead load ratio. Shear design did not cause any problem due to
the high ratio of dead to live load, and mostly symmetrical total
load.

A very significant characteristic of such structural system

is that adding a very small prestressing force can significantly
affect the deformation while, at the same time, a small change of
internal forces takes place in the system. By prestressing the
lower chord, the upper reinforced concrete chord is elastically
supported. The stiffness of the supports depends on both the
tensile capacity and the configuration of tendons. Therefore, the
deflection of the main girders can be easily controlled by the
suitable choice of prestressing force. As for the time-dependent
deflection of such systems, it is not considerably greater than
the elastic one, even for very high creep and shrinkage, as it is
also primarily governed by the shape and the deformation of the
lower chord tendons [ 4]. In the case of this hangar, the expec-
ted average creep factor is about 2, due to the construction techno-
logy.
 Prestressing 199

The analysis has shown that, due to the creep and shrin-
kage of concrete, forces in such systems change very little with
time. This explains why the loss of prestressing force with time,
due to the time-dependent deformations of concrete,is incomparably
lower than in classically prestressed concrete girder with tendons
inside the cross-section.

The main girders have been cast on scaffolds, right under

the final position, about 9 m above the floor of the hangar, which
was the height necessary to provide the tendons rise. The pres-
tressing of the main girders was designed to be carried out in
two stages, simultaneously from both ends. The first stage of
prestressing was performed while the girders were still on scaf-
folds and resulted in 15 em camber. At that moment, the girders,
were loaded only by the self-weight and the load of suspended
short secondary girders and purlins which were designed to be
lifted together with the main girder, as shown in Fig. 6 and 7 .
The prestressing force of the first stage was 27. 200 kN, making
56% of the allowed value. It produced 15 em camber at the midspan.

After the first stage of prestressing was completed, the

main girders were lifted for about 24 m to the designed positions
on the top of the main columns. The lifting weight of the main
girders, together with the hanged short secondary girders and
purlins, was 1, 700 t each.

Fig. 7 shows the lifting of the main girders. Two main

girders are already in place on the top of the columns, while the
last one is lifted halfway.

The second and final stage of prestressing was performed

after the application of the total dead load, with the exception of
thermo- and hydro-insulation. The final prestressing force of
37,500 kN, making 78% of the allowed value, was determined to
provide the 15 em camber of the main girders at the midspan,
with respect to the final designed position. As the loss of the
prestressing force in already prestressed tendons due to the
elastic deformations in such structural systems can exceed 50%,
two or even more cycles of tensioning were necessary to provide
the designed forces in all tendons. The designed extreme total
force in all 9 tendons amounts to 41,800 kN, or about 87% of the
admissible value. After the second stage of prestressing, the
tendons were injected by simple cement emulsion.

The secondary mountable roof girders, shown in Fig. 8,

are also two-chord structural systems, with a span of 15.65 m
and rise of 1. 50 m. The upper, compressed chord is of reinforced
concrete while the lower chord and the two vertical chairs are
made of rigid steel box sections. Like the main girders, the
secondary roof girders are very rational, economical and relatively
light-weight.
200 Ivkovic and Perisic

CONCLUSIONS

The design and construction of the new prestressed con-

crete hangar at the Belgrade International Airport, using the
described structural ideas and systems, have confirmed that com-
posite two-chord systems with prestressing steel elements outside
the compressed reinforced concrete chord are very suitable and
very economical structures for roofs of large and exceptionally
large spans. Contrary to the former experience, at least in Yugo-
slavia, this structure showed significant economical and technical
advantages compared with the corresponding steel or classically
prestressed concrete design alternatives. Using the numerous
advantages of such structural systems, a relatively light,
mountable and economical roof structure, and especially a very
short term of construction have been obtained for this hangar.
With the structural system of the main girders outside the covered
part of the building, free in space but requiring no special main-
tenance during the service life, and with the roof structure
hanged on the main girders, the minimum effective volume and
the minimum facade surfaces, as well as the minimum service and
maintenance costs have also been achieved.

REFERENCES

1. Bazant, Z. P. , "Prediction of Concrete Creep Effects Using

Age Adjusted Effective Modulus Method", ACI Journal, Vol. 69,
(1972)' pp. 212-217.
2. Ivkovic, M. , Acic, M. , Perisic, Z. , Pakvor, A. , "Concrete
Structures with Steel Elements Outside the Concrete Section,
Proceedings of the 12th IABSE Congress, Vancouver, 1984.
3. Ivkovic, M. , Acic, M. , Perisic, Z. , Pakvor, A. , "Demountable
Concrete Structures with Steel Elements Outside the Concrete
Section", Proceedings of the International Symposium "Demoun-
table Concrete Structures", Rotterdam, 1985.
4. Ivkovic, M., Perisic, Z., Pakvor, A., Acic, M., "New Pres-
tressed Concrete Hangar at the Belgrade International Airport
in Yugoslavia, Proceedings of the lOth FIP Congress, New
Delhi, 1986.
5. Volterra, V. , "Theory of Functlonals and Integral and
Integrodifferential Equations", Dover, New York, 1959.
 Prestressing 201

Fig. 1--Elernents of the two-chord structural system

202 lvkovic and Perisic

Fig. 2--Differential element of the two-chord system

 Prestressing 203

P,{x)
lm:;w;IJI

Fig. 3--Example of the two-chord structural system with the

prestressing tendons outside the concrete cross section
 1L F
m ~~ f
~

~ -....
<
~

~
()

§
0...
~
::::!.
....
a>
()

~ ~~ ~ 18.~ ~ l
i 25.90

m
B
50.40 25.90

Fig. 4--The general plan of the new prestressed concrete hangar at the Belgrade International
Airport in Yugoslavia
 Prestressing 205

AriD
AL~--------------------,~j7·~~--------------------T

A-A B-B c-c

¢ :-1
l!l25~5

I
Cl
Fig. 5--The view and characteristic cross-sections of the main girder

Fig. 6--The main girder under construction

206 lvkovic and Perisic

Fig. 7--Lifting of the main girders: first two girders are already
lifted while the third one is half-way up

Fig.8--The secondary roof girders

 SP 120-10

Strengthening of Existing Bridges

(Simple and Continuous Span)
by Post-Tensioning
by EW. Klaiber, K.E Dunker,
and W.W. Sanders, Jr.

Synopsis: Approximately 40% of the bridges in the United States

are classified as deficient and in need of rehabilitation or
replacement. Of these bridges, many are classified as deficient
because their load-carrying capacity is inadequate for today' s
increased traffic. This insufficient load-carrying capacity has
resulted from poor maintenance, increases in legal load limits,
deck overlays, changes in design specifications, and other
factors. In response to the need for a simple, efficient
procedure for strengthening existing bridges, the authors have
been investigating the use of post-tensioning. The authors have
investigated the use of post-tensioning on simple span bridges as
well as on continuous span bridges. Various post-tensioning
schemes have been tested on laboratory models as well as on
existing bridges. This paper will briefly review the post-
tensioning research that has been completed by the authors in the
past few years. This work indicates that post-tensioning is a
viable, economical technique for flexural strengthening of steel-
beam composite-concrete deck bridges.

Keywords: bridges Cstn1ctures); flexural strength; loads

(forces); post-tensioning; renovating; research; spans;
strengthening; stresses; stnlctural design; tests; unbonded
prestressing

207
208 Klaiber, Dunker, and Sanders, Jr.
F. Wayne Klaiber, FACI, is Professor of Civil and Construction
Engineering and Manager of the Bridge Engineering Center at Iowa
State University, Ames, Iowa. He is a member of ACI Committee
215, Fatigue of Concrete and ACI Educational Committee, E801.

Kenneth F. Dunker, also an ACI Member, is an Associate Professor

of Civil and Construction Engineering at Iowa State University.
He is a Professional Engineer and has been involved in bridge
strengthening research since 1980.

Wallace W. Sanders, Jr, is Associate Dean for Engineering Research

at Iowa State University and has conducted research on design and
behavior of bridges for over 30 years. Dr. Sanders has conducted
studies on load distribution, strengthening and behavior of steel,
concrete and timber railway and highway bridges.

INTRODUCTION

About one-half of the approximately 600,000 highway bridges

in the United States were built before 1940, and many have not
been adequately maintained. Most of these bridges were designed
for lower traffic volumes, smaller vehicles, slower speeds, and
lighter loads than are common today. In addition, deterioration
due to environmental factors is a growing problem. According to
the Federal Highway Administration (FHWA), almost 40% of the
nation's bridges are classified as deficient and in need of
rehabilitation or replacement. Many of these bridges are
deficient because their load-carrying capacity is inadequate for
today' s traffic. Strengthening can often be used as a cost-
effective alternative to replacement or posting for restricted
loads.

For the past eight years, the authors have been investigating
the use of external post-tensioning for strengthening existing
bridges. Although this work involves numerous projects, for
discussion purposes it will be presented here as follows:

Project 1. Strengthening of Existing Single-Span Steel-

Beam Concrete Deck Bridges: Phases I, II, and
III.

Project 2. Strengthening of Existing Continuous Composite

Bridges: Phases I and II.

This paper will review the purpose and results of these

research projects. Project 1 and Phase I of Project 2 are
completed, so final results can be reported. Because Phase II of
Project 2 is still in progress, only preliminary results can be
presented.
 Prestressing 209
PROJECT 1

As previously noted, Project 1 consisted of three separate

phases. Portions of Project 1 have been documented in the
literature (1,2); thus, it will be reported here only briefly.
Phase I of the research program (3) was to determine a method of
strengthening single-span, composite ste8l-beam concrete deck
bridges. Primarily it was the exterior beams of these bridges
which required strengthening.

After reviewing several strengthening techniques, we found

that post-tensioning the exterior beams with high-strength tendons
was the most promising technique. The feasibility of this
strengthening technique was verified analytically, by using
orthotropic plate theory, and experimentally, through testing of
various post-tensioning schemes on a half-scale model bridge (L =
26' -0", W = 15' -8 3/8") shown in Fig. 1. Although various post-
tensioning schemes were investigated (post-tensioning of all
beams, post-tensioning of exterior beams only, post-tensioning of
one exterior beam, etc.) to obtain data on the lateral
distribution in the bridge, only the exterior beams actually
required post-tensioning to reduce the overstress. However, when
the exterior beams were subjected to post-tensioning, due to
lateral distribution, significant stress reduction resulted in the
interior beams. The midspan bottom flange strains that resulted
from subjecting the external beams to a post-tensioning force of
20 kips are shown in Fig. 2. Excellent agreement can be seen
between the measured strains and the theoretical curve.

As a result of the success of the Phase I laboratory

investigation, Phase II (4,5) of the investigation was undertaken.
Phase II consisted of field strengthening and monitoring two
existing bridges. One of the bridges, located in northwestern
Iowa (Bridge 1: L =52', W = 31'-10 1/2"), was a prototype of the
model bridge tested during Phase I. The other bridge, located in
central Iowa (Bridge 2: L 72', W 31'-10 1/2"), was
considerably longer and had a 45° skew.

Photographs of the post-tensioning brackets used on the two

bridges are shown in Fig. 3. The brackets used on Bridge 1 (Fig.
3a) were similar to the brackets used on the model bridge. Figure
3 shows two tendons (1 1/4" in diameter) used on Bridge 1, while
four tendons (2 - 1/2" in diameter and 2 - 1" in diameter) were
used on Bridge 2. Although only the exterior beams of both
bridges were post-tensioned, because of lateral distribution
significant stress reduction was also obtained in the interior
beams.

For composite action, both bridges had angle-plus-bar type

shear connectors. Analysis indicated that for the bridges to
carry additional live load, the exterior beams of Bridge 1 and all
the beams of Bridge 2 needed additional shear connectors. An
assumption- -that the type of steel in some bridges requiring
strengthening might be unknown- -prompted the decision to add shear
210 Klaiber, Dunker, and Sanders, Jr.
connectors by bolting rather than by welding. High-stre~gth bolts
tested as shear connectors in the laboratory were found to be
somewhat stronger than shear studs of equivalent size (6). In the
field, cores (4" in diameter) were drilled in the bridge deck
above the beams in the two bridges, and high-strength bolts were
double-nutted to the beams. Later the cores were filled with
nonshrink grout.

Both bridges were instrumented to determine strains and

deflections. To determine the bridges' response to post-tension
strengthening, they were both load tested with overloaded trucks
(see Fig. 4 for details) before and after the exterior beams were
post-tensioned. The two bridges were initially strengthened and
tested during summer 1982 and were retested during summer 1984.
During the intervening time period, the bridges were periodically
inspected; no behavior changes were noted in either bridge.

Bridge 1: During the original post-tensioning of Bridge 1, each

exterior beam was subjected to an average force of 182.0 kips, to
produce a strain reduction of 218 ~€ in the lower flange
coverplates of the exterior beams. However, because of the
presence of end restraint, a strain reduction of only 145 ~€ was
obtained. When the post-tensioning force was removed from the
bridge during summer 1984, a magnitude of 172.0 kips was measured.
This decrease in force (5.5%) is primarily the result of reduction
in the amount of end restraint present, however, a small
percentage of the decrease is due to relaxation of the post-
tensioning tendons. A post-tensioning force of 196.8 kips, which
produced an average upward displacement of 0.18 in. in the
exterior beams, was reapplied to each exterior beam during summer,
1984. The midspan bottom flange coverplate strains resulting from
this post-tensioning force are shown in Fig. 5. Also shown in the
figure are theoretical curves, obtained from finite element
analysis, for both simple supports and fixed ends. The measured
strains are essentially midway between the two theoretical curves,
thus indicating the presence of significant end restraint.
Additional verification of the presence of end restraint is the
sizable magnitude of strains measured at instrumented sections 9
in. from the centerline of support. Although the desired strain
reduction was not obtained, end restraint was also found when live
load was placed on the bridge. Thus, end restraint reduced the
effectiveness of the post-tensioning and also reduced the strains
produced by live load.

Bridge 2: Initially, each exterior beam was subjected to an

average post-tensioning force of 305.6 kips that, based on the
assumption of simple supports, was to produce a strain reduction
in the lower flange of 212 ~€. However, when this force was
applied, a reduction of slightly less than half that amount (102
~€) was obtained because of the skew effect and the presence of
end restraint. During summer, 1983, this bridge was resurfaced
and had major portions of the curb sections removed and replaced.
Because the post-tensioning force was still applied to the bridge,
large upward displacements were visible upon removal of the dead
 Prestressing 211
load and reduction in the effective section. When the post-
tensioning force was removed from the bridge during summer, 1984,
a magnitude of 271.5 kips was measured. This decrease of 11.2% is
thought to be primarily the result of losses incurred during
rehabilitation of the bridge which took place the previous summer.
A post-tensioning force of 371.2 kips was then reapplied to each
exterior beam, resulting in an upward displacement of 0.33 inches.
Shown in Fig. 6, along with the theoretical curves assuming simple
supports and fixed ends, are the midspan bottom flange coverplate
strains resulting from this post-tensioning force. The measured
strains on the exterior beams are closer to the fixed end
theoretical curves than to the simple span theoretical curves.
Because of the skew, the end restraint present in this bridge was
greater than that observed in Bridge 1. As in Bridge 1,
significant strains were measured in the vicinity of the supports;
these also verify the presence of end restraint. End restraint
again reduced the effects of live load. In effect, the live load
tension strains, which were lower-than-expected, compensated for
the lower-than-expected post-tensioning compression strains.

The third phase of the investigation (7) involved the

development of a simplified design methodology for practicing
engineers so that they could avoid use of orthotropic plate theory
and/or finite-element analysis.

By analyzing an under-capacity bridge in which the exterior

beams are smaller than the interior beams, an engineer can
determine the overstress in the exterior beams. This overstress
is based on the procedure of isolating each bridge beam from the
total structure. The cross-sectional properties of the individual
beams are based on the rules given in the AASHTO Standard
Specifications for Highway Bridges (8). The amount of post-
tensioning required to reduce the stress in the exterior beams can
then be determined if the amount of post-tensioning force
remaining on the exterior beams is known. The amount of force
remaining on the exterior beams can be quantified with force and
moment fractions. A force fraction, FF, is the ratio of the axial
force that remains on a post-tensioned beam at midspan to the sum
of the axial forces for all bridge beams at midspan, while a
moment fraction, MF, is simply the moment remaining on the post-
tensioned beam divided by the sum of mid-span moments for all
bridge beams. Knowing these fractions, the required post-
tensioning force may be determined by utilizing the following
relationship:

f = FF [ _i_ ] + MF [ Pee ] (1)

A I

where f desired stress reduction in lower beam flange.

p post-tensioning force required on each exterior
beam.
A cross-sectional area of exterior beams.
e = eccentricity of post-tensioning force measured
from a bridge's neutral axis.
212 Klaiber, Dunker, and Sanders, Jr.
c = distance from neutral axis of beam to lower
flange.
I moment of inertia of exterior beam at section
being analyzed.

Distribution factors were determined by a finite element analysis;

the accuracy of the finite element model was determined through
use of the data obtained in the laboratory and field testing. A
series of experiments was conducted with the finite element model
to determine the sensitivity of the distribution factors to the
various bridge variables. The possibility of having one
distribution factor for axial force and moment was investigated;
however, it was found that a much greater amount of axial force
than moment remains on the exterior beams at midspan. Thus for
design purposes, force fractions, FF, and moment fractions, MF,
should be kept separate as shown in Eqn. 1. These experiments
demonstrated that variables such as the stiffness of shear
connectors, cover plate length, bridge skew, and deck crown had
no significant effects. Those variables having significant
effects on the distribution factors are discussed briefly in the
following paragraphs.

End Restraint: As previously noted, a considerable amount of end

restraint was found in the two actual bridges strengthened.
Through analysis it was found that end restraint increased the
amount of post-tensioning force and moment retained by the
exterior beams. However, because of the difficulty in quantifying
end restraint existing in actual bridges, partial end restraint
was neglected in the design methodology. This is conservative in
that simple support distribution factors are smaller than those
for actual support conditions (i.e., partially restrained ends).

Span and Anchorage Location: Span length was found to be the most
significant variable in the moment fractions; exterior beams were
found to retain less moment as the span length was increased.
When the span length was held constant and anchorages were moved
toward the supports, more moment was distributed away from the
exterior beams. The distribution is thus dependent on the length
of the beam subjected to post-tensioning (i.e., distance between
anchorages).

Relative Beam Stiffness: Moment distribution was found to be

dependent on relative beam stiffness; stiffer exterior beams
retained more post-tensioning moment. Thus, if the exterior beams
were smaller than the interior beams, the exterior beams retained
considerably less of the post-tensioning moment than if the beams
were all the same size.

Utilizing the finite element model, bridges having various

combinations of the involved variables were analyzed to determine
the midspan force and moment fractions. By using a multiple
linear regression procedure, the calculated fractions were
analyzed to identify the variables that had the most effect on the
distribution fractions. Relatively simple equations with high
 Prestressing 213
degrees of accuracy were derived for the distribution fractions.
For the distribution fractions at midspan and at other locations
within the span, the reader is referred to Refs. 2 and 7.
Reference 7 presents a complete design methodology in which other
considerations such as gain in post-tensioning force due to bridge
deflections resulting from live load, loss of post-tensioning
force due to temperature differentials, and the like are
considered.

The strengthening procedure and design methodology just

described have since been used on several bridges in the states of
Iowa, Florida, and South Dakota. In all instances, the procedure
was employed by local contractors without any significant
difficulties.

PROJECT 2

Phase I: As previously noted, only Phase I (9,10,11) of this

project is completed while Phase II is currently in progress.
Thus, final results of Phase I are presented, while only
preliminary findings of Phase II are given. In addition to the
problems with single-span bridges that were addressed in Project
1, a large inventory of continuous composite bridges are also
structurally inadequate. Some of these bridges are overstressed
in the negative moment regions, some in the positive moment
region, while a large number are overstressed in both regions.
On the basis of Project 1 results, it appeared that post-
tensioning could also be used to strengthen continuous bridges.
Thus, the primary objective of this phase of the investigation was
to determine the feasibility of strengthening continuous composite
bridges by post-tensioning. More specific objectives of the study
were to determine:

• the distribution of post-tensioning forces and

moments in the positive and negative moment regions

• the best tendon arrangement (straight, inclined,

continuous, etc.) for post-tensioning the positive
and negative moment regions

• the contribution of the deck to composite action

in negative moment regions.

The objectives of this project were fulfilled through testing of

a one-third-scale model bridge, testing of a full-scale mockup of
the negative moment region of a bridge beam, and the development
of a finite element model. The model bridge, mockup, testing
program, and some representative results will be discussed in the
following paragraphs. The finite element model developed was
calibrated and verified by using results from the laboratory model
tests. Thi~ program was used to investigate the effect of the
various bridge parameters on post-tensioning force and moment
distribution.
214 Klaiber, Dunker, and Sanders, Jr.
Because of space limitations in the Iowa State University
(ISU) Structural Engineering Laboratory, it was necessary to limit
the bridge model in number of spans as well as scale. After
consideration of various combinations, we decided to construct a
four-beam, three-span continuous bridge. An overview of the model
bridge is given in Fig. 7. The overall dimensions of the bridge
are total length= 41'-11", width= 8'-8", deck thickness =·2.25",
and individual span lengths= 12'-8", 16'-4", 12'-8". To solve
the problem of dead load on the model and to prevent uplift when
the model was subjected to certain loading patterns, each beam was
vertically restrained at each support. The system employed (see
Fig. 6) prevented vertical movement; however, it did not restrict
horizontal movements. To model accurately the one- third- scale
bridge beams, the beam sections were fabricated from steel sheet
and plate. The exterior and interior beams were 6" and 7" in
depth, respectively.

As may be seen in Fig. 7, the post-tensioning in the negative

moment regions was applied on top of the deck surface. This was
done for convenience; obviously, in the field, in the negative
moment region the post-tensioning would have to be done under the
deck, or a portion of the deck would have to be removed to
accommodate the post- tensioning sys terns and then be replaced
later.

The full-scale mockup of the negative moment region

previously noted was constructed so that the strength and behavior
of post-tensioning arrangements correctly positioned under the
deck (which obviously was not possible on the bridge model) could
be determined. The mockup was strengthened by using both straight
and harped post-tensioning tendon profiles. Various post-
tensioning schemes were tested on the mockup with and without the
presence of vertical loading. Space limitations preclude
additional discussion of the mockup tests herein. However, those
desiring additional information are referred to Ref. 11.

A total of 66 strain gages were mounted at critical sections

of the model bridge beams; each tendon was also instrumented with
strain gages so that post-tensioning forces could be accurately
measured. Deflections at the midspan of each beam (twelve
displacements) were measured by using direct current differential
transformers. To facilitate rapid collection of data, all strains
and displacements were measured and recorded with a data
acquisition system.

The test program employed consisted of determining the

bridge's response, strains, and deflections to the following
loading conditions:

• vertical concentrated load at various locations

on the middle and end spans

• various arrangements of post-tensioning (i.e.,

negative moment regions of exterior beams,
 Prestressing 215
negative moment regions of all beams, positive
moment regions of end spans only, etc.)

• combination of the various post-tensioning

arrangements plus vertical loading.

The following paragraphs discuss briefly some of the results

from this investigation. Post-tensioning continuous bridges is
considerably more complex than post-tensioning simple-span bridges
in that longitudinal and transverse distribution effects both must
be determined. Twenty-two different post-tensioning schemes
(symmetrical and asymmetrical) were tested on the bridge model.

Shown in Fig. 8 are the measured strains for an asymmetrical

post-tensioning scheme (positive moment regions of exterior Beam
4 subjected to a nominal post-tensioning force of 20 kips). This
scheme of post-tensioning could be utilized to strengthen an
exterior beam that has been damaged by an overheight vehicle or
weakened by localized corrosion. In the beam post-tensioned, Beam
4, the strains are opposite of those resulting from dead load.
Strains in the bottom flange are tensile at midspans and
compressive ac the interior supports. Thus, these strains reduce
dead load strains at critical locations, thereby increasing the
live load capacity of the beam. Transverse distribution effects,
similar to those in single-span bridges, transfer beneficial
strain reductions to the adjacent beam, Beam 3 (See Fig. 8b).
Post-tensioning Beam 4 has a very small effect on Beams 1 and 2
(see Fig. Be and 8d); however, the effect on Beam 1 is adverse in
that the strains produced in Beam 1 by post-tensioning Beam 4 are
additive to the dead load strains in Beam 1. Other asymmetrical
tests produced similar results; post-tensioning a specific region
of a beam created beneficial strains at that region; however, it
also produced adverse strains in other regions. Therefore,
asymmetrical post-tensioning, due to distribution effects, should
be thoroughly investigated before being employed. to strengthen a
particular region of a bridge.

Shown in Fig. 9 are the experimental and theoretical bottom

flange strains for two symmetrical load cases. The theoretical
curves shown in this figure were obtained by using the finite
element analysis previously discussed. In Case 1 (see Fig. 9a)
the positive moment regions of all exterior beams are post-
tensioned, while in Case 2 (see Fig. 9b) the negative moment
regions of all exterior beams are post-tensioned. In both cases
a nominal force of 20 kips was applied to each beam. In both of
these cases there is excellent agreement between the measured
strains and the theoretical curves.

As previously noted in both cases, only the exterior beams

were post-tensioned. Thus, the strains that are shown in the
interior beams are the result of lateral load distribution. As
this figure indicates, distributions are approximately the same,
in terms of distribution fractions, for both positive moment
region post-tensioning and negative moment post-tensioning.
216 Klaiber, Dunker, and Sanders, Jr.
Also shown in Fig. 9 are the longitudinal strain
distributions. Both post-tensioning schemes produced beneficial
strains (i.e., opposite to those produced by dead load) both at
midspan and at interior supports. In Case 1, the positive moment
regions were post-tensioned, producing compressive strains in the
bottom flange near midspan and tensile strains in bottom flange
in the vicinity of the interior supports. Similar results (see
Fig. 9b) were obtained when the negative moment regions of the
exterior beams were post- tensioned. Thus, depending on the
bridge, it may be possible to strengthen it (both in positive and
negative moment regions) by post-tensioning only the positive
moment regions of the exterior beams or the positive moment
regions of all beams. Post-tensioning the positive moment regions
is obviously simpler than post-tensioning the negative moment
regions of a given bridge because of the presence of the deck.

In the strengthening of a continuous span bridge by post-

tensioning, the engineer has so many variables (beams to be post-
tensioned, region(s) of beam to be post-tensioned, magnitude of
post-tensioning force, post-tensioning length, etc.) that in
essentially all cases it is possible to "fine-tune" the bridge
(i.e., reduce strains in the desired locations by the desired
amount).

Phase II: Phase II of this investigation was undertaken as a

result of the successful laboratory investigation (Phase I). The
objectives of this phase of the investigation were

• to design and install a post-tensioning strengthening

system for a selected continuous span, steel beam-
composite concrete-deck bridge

• to instrument and test the bridge for determination

of its behavior before and after post-tensioning

• to monitor the bridge's behavior for an extended

period of time.

The bridge selected for strengthening (see Fig. 10) is

located in northwest Iowa, near the town of Fonda. The four-beam
bridge consists of three spans (45' -9", 58' -6", 45' -9"); the
exterior beams (W2lx62) and the interior beams (W24x76) bear at
approximately the same elevation. The difference in beam heights
provides a crown in the bridge deck (nominal deck thickness 6
3/4"). The similarity between this bridge and the laboratory
model will permit comparisons of field and laboratory data.

On the basis of an analysis of the bridge, it was determined

that the desired stress reduction in the bridge (in both the
positive and negative moment regions) could be accomplished by
post-tensioning the positive moment regions of all beams in the
bridge (12 different regions). Shown in Fig. 11 (upper number in
box) are the theoretical post-tensioning forces required on each
beam to obtain the desired stress reduction at the various
 Prestressing 217
overstressed regions in the bridge. The lower number in the box
indicates the magnitude of force that was actually applied to each
beam. In all but two beams slightly more post-tensioning force
was applied than was theoretically required.

The bridge was instrumented with 66 weldable strain gages

positioned at critical sections of the beams; each of the 24
tendons (two required per beam) were instrumented so that post-
tensioning forces could be accurately determined. Deflection at
the midspan of each beam in one end span, each beam in the middle
span, and one exterior beam and one interior beam of the other end
span (10 displacements) were measured by using direct current
differential transformers. A data acquisition system facilitated
rapid data collection.

The testing program utilized on this bridge was essentially

the same as that used on the single span bridges (Bridges 1 and
2). Data were recorded when the bridge was subjected to an
overloaded truck (total weight= 57.92 kips) positioned at various
predetermined locations before and after post-tensioning, and at
various stages of the post-tensioning sequence.

As the testing program previously described was only recently

completed, analysis of the data obtained is still in progress. A
preliminary review of the data indicates that the bridge's
behavior is essentially the same as the laboratory model's except
for the presence of end restraint found in the field.

This bridge will be inspected approximately every three

months for one year to monitor its behavior. At the end of that
period the bridge will again be service-load tested to determine
any behavioral changes from its initial strengthening.

SUMMARY

The results of the research outlined in this paper indicate

that post-tensioning is a viable, economical technique for
flexurally strengthening simple span and continuous span steel-
beam and concrete-deck composite bridges. The post-tensioning
systems employed on the single span bridges and the continuous
bridges essentially reduce dead load strains (stresses) in the
bridge·s, thus making it possible for the bridges to carry
additional live loads. Post-tensioning will slightly increase the
ultimate strength of the bridge and in most cases induce a small
upward camber. It does not, however, significantly reduce live
load deflections or significantly influence live load
distribution.

ACKNOWLEDGMENTS

The research reviewed in this paper was conducted by the

Bridge Engineering Center, Engineering Research Institute of Iowa
218 Klaiber, Dunker, and Sanders, Jr.
State University and funded by the Iowa Highway Research Board and
the Highway Division, Iowa Department of Transportation, Ames,
Iowa. The authors thank the engineers at the Iowa DOT, especially
William A. Lundquist, Bridge Engineer, and John P. Harkin, Bridge
Rating Engineer, for their support, cooperation and counsel.
Special thanks are accorded the numerous graduate students who
assisted with the various phases of the two projects. The
opinions, findings, and conclusions expressed herein are those of
the authors and not necessarily those of the Iowa Department of
Transportation or the Iowa Highway Research Board.

REFERENCES

1. Klaiber, F. W.; Dunker, K. F.; and Sanders, W. W. Jr.

"Strengthening of Single- Span Steel-Beam Bridges." Journal
of the Structural Division, ASCE, V. 108, No. 12, Dec. 1982,
pp. 2766-2780.

2. Dunker, K. F.; Klaiber, F. W.; and Sanders, W.W., "Jr.

"Post-Tensioning Distribution in Composite Bridges."
Journal of Structural Engineering, ASCE, V. 112, No. 11,
Nov. 1986, pp. 2540-2553.

3. Klaiber, F. W.; Dunker, K. F.; and Sanders, W. W. Jr.

"Feasibility Study of Strengthening Existing Single Span
Steel Beam Concrete Deck Bridges." Final Report. ERI
Project 1460. ISU-ERI-Ames-81251, Engineering Research
Institute, Iowa State University, June 1981, Ames, Iowa, 141
pp.

4. Klaiber, F. W.; Dedic, D. J.; Dunker, K. F.; and Sanders,

W. W. Jr. "Strengthening of Existing Single Span Steel Beam
and Concrete Deck Bridges." Final Report--Part I, ERI
Project 1536, ISU-ERI-Ames-83185, Engineering Research
Institute, Iowa State University, Feb. 1983, Ames, Iowa, 185
pp.

5. Dunker, K. F.; Klaiber, F. W.; Beck, B. L.; and Sanders, W.

W. Jr. "Strengthening of Existing Single-Span Steel-Beam
and Concrete Deck Bridges." Final Report- -Part II. ERI
Project 1536. ISU-ERI-Ames-85231, Engineering Research
Institute, Iowa State University, March, 1985, Ames, Iowa,
146 pp.

6. Dedic, D. J.; and Klaiber, F. W. "High-Strength Bolts as

Shear Connectors in Rehabilitation Work." Concrete
International: Design and Construction, ACI, V. 6, No. 7,
July, 1984, pp. 41-46.

7. Dunker, K. F.; Klaiber, F. W.; and Sanders, W. W. Jr.

"Design Manual for Strengthening Single-Span Composite
Bridges by Post-Tensioning." Final Report--Part III, ERI
Project 1536, ISU-ERI-Ames-85229, Engineering Research
 Prestressing 219
Institute, Iowa State University, March, 1985, Ames, Iowa,
102 pp.

8. American Association of State Highway and Transportation

Officials. Standard Specifications for Highway Bridges.
13th Edition. Washington: American Association of State
Highway and Transportation Officials. 1983.

9. Dunker, K. F.; Klaiber, F. W.; Daoud, F. K.; Wiley, W.E;

and Sanders, W. W. Jr. "Strengthening of Existing
Continuous Composite Bridges." Final Report, ERI Project
1846. ISU-ERI-Ames-88007, Engineering Research Institute,
Iowa State University, July, 1987, Ames, Iowa, 220 pp.

10. Daoud, F. K. "Experimental Strengthening of a Three-Span

Composite Model Bridge by Post-Tensioning." Thesis
presented to Iowa State University, at Ames, Iowa, in
partial fulfillment of the requirements for the degree of
Master of Science, 1987, 140 pp.

11. Wiley, W. E. "Post-Tensioning of Composite T-Beams

Subjected to Negative Moment." Thesis presented to Iowa
State University, at Ames, Iowa, in partial fulfillment of
the requirements for the degree of Master of Science, 1988,
227 pp.
220 Klaiber, Dunker, and Sanders, Jr.

Fig. 1. Single Span Composite Bridge Model

(Bridge Model No. 1).
 Prestressing 221

(a) Bridge No. 1.

(b) Bridge No. 2.

Fig . 3. Post-Tensioning Brackets and Tendon Arrangements.

222 Klaiber, Dunker, and Sanders, Jr.

14'- 8"

AXLE LOADS: 13.64kips 46.90kips

TOTAL: 60.S4kips

a. Truck used on bridge 1.

4'- 3"

AXLE LOADS: 13.36kips 23.57kips 23.S7kips

TOTAL: 60.SOkips

b, Truck used on bridge 2.

Fig. 4. Field Test Vehicles Used on Bridges 1 and 2.

 Prestressing 223

.
0
FIXED ;_N~---""'
:I:
u
~
0::
w -80
\
//
/
/
/
..' ' .....................

'
"-
II\
w
:I:
u
~
0
0::
~
:::!:
z
<
0::
1-
II\
e FIELD MEASUREMENT

BM1 8M2

Fig. 5. Post-Tensioning, Midspan, Bottom Flange Beam Strains

(Bridge No. 1).

. .
~~IMPlE ~PA?
u
:I:

~ //
/ ''
0:: -80 / ''
w
"-
/
/
'
II\
w
:I:
u
• •
z
0
0::
~ SIMPLE SPAN
:::!:
z
<
0::
1-
II\
e FIELD MEASUREMENT
-320
BM1 MB2 8M3 BM4

Fig. 6. Post-Tensioning, Midspan, Bottom Flange Beam Strains

(Bridge No. 2).
224 Klaiber, Dunker, and Sanders, Jr.

Fig. 7. Continuous Span Composite Bridge Model

(Bridge Model No. 2).
 Prestressing 225

500,.....
4. ~~
0

-500 • •
•
L-

a. Beam 4

,.....
500

u
:I:
z 0
,. tt
cc
1.1.1
Q.. • • •
""
1.1.1
:I:
-500 '--
b. Beam 3
u
z
0
cc ,.....
~ 500
:::!:
z 0 ...
<(
cc • •
....
"" -500 L-
c. Beam 2

500,.....

0 • • •
d. Beam 1
-500 .__
20.20kips 19.71kips 19.92kips

~I I I l
Beam and Load Key

Fig. 8. Bottom Flange Beam Strains from Post-Tensioning

Positive Moment Regions of Beam 4 (Bridge Model No. 2).
226 Klaiber, Dunker, and Sanders, Jr.

:I:
u
~
a:
UJ
0..
Ill
UJ
:I:
u
~
0
a:
u
:E

EXTERIOR BEAM

e EXPERIMENTAL STRAIN

a. Case 1: Positive moment regions of exterior beams post-

tensioned.

:I:
u
~
a:
UJ
25~[
•
0..
Ill
UJ
:I:
-250 ---- INTERIOR BEAM
u
~
500
I
0
a: I
u
:E 0
z"
<
a:
-500
1-
Ill -1000

b. Case 2: Negative moment regions of exterior beams post-

tensioned.

1- 1- MIDSPA~
1I· I-MIDSPAN
I
5PAN1 =12'-8" ~
MIDSPAN

A•I•
SPAN 2 I= 16' -...:.4'_'_ _r._ .:.S:. .:PA.:.N.:. .3:._~. .:.1.:. .2'_·.:.8'_'_:A-.!I
-·

Fig. 9. Bottom Flange Beams Strains (Bridge Model No. 2).

 Prestressing 227

Fig. 10. Fielding Testing of Continuous Bridge.

228 Klaiber, Dunker, and Sanders, Jr.

[I[]
117 [TIE
143 cru
123

lliiJ
115 [ill]
167 [;]
105

ffiB
102 [ill]
176 [I[]
114

[ili]
122 [ill]
148 illf]
112

B= REQUIRED FORCE, kips

ACTUALLY APPLIED FORCE, kips

Fig. 11. Post-Tensioning Forces Applied to Continuous Bridge.

 SP 120-11

Post-Tensioned Repair and Field Testing

of a Prestressed Concrete Box Beam Bridge
by A.E.N. Osborn and H.K. Preston

ABSTRACT

Over 600 bridges composed of adjacent prestressed concrete box

beams were built in the early 1950's in Pennsylvania. The box
beams were placed side-by-side and had an asphalt wearing course
on top. Span lengths ranged from 30 to 60 feet. Their design was
very conservative by today's standards.
In these bridges, the concretes have high chloride contents; water
leaks down through the joint between box units and the strands
often have inadequate cover. Thus, it is not surprising that many
of the box beams are deteriorating due to corrosion of their
prestressing strands.
This project was directed toward developing economical repair
schemes for these bridges. The literature survey did not reveal
any schemes specifically applicable to adjacent box beams.
Two external reinforcement repair schemes were developed, and
trial installations were made on a bridge near York, Pennsylvania.
Both schemes included the removal of deteriorated concrete,
placement of external reinforcement beneath the beam and
application of gunite to the soffit of the beam.
In Scheme 1, the external reinforcement consisted of epoxy-coated
reinforcing bars. This repair method restored ultimate flexural
capacity, but did not restore lost prestress. It was the least
costly of the two methods.
In Scheme 2, post-tensioned, epoxy-coated strand were used. This
restored the full ultimate flexural capacity and most of the lost
prestress. Difficulties were encountered in installing anchors
for the post-tensioned system, but its performance was good.
The bridge was tested after repair. The external reinforcements
were found to be fully composite with the original beams. The
tests also revealed the lateral distribution of wheel loads. In
spite of the poor condition of the bridge, the wheel loads were
well distributed laterally, leading to a structure which was
stronger and stiffer than expected.
Keywords: box beams; bridges (structures); coatings;
deterioration; epoxy resins; lateral pressure; loads (forces);
load tests (structural); post-tensioning; prestressing steels;
repairs; shotcrete

229
230 Osborn and Preston

Andrew E. N. Osborn and H. Kent Preston

Wiss, Janney, Elstner Associates, Inc., Princeton Junction, NJ

INTRODUCTION

Numerous prestressed concrete bridge structures were constructed

in Pennsylvania in the infancy of the prestress industry. Several
fabrication practices used at that time have proven to lack
durability. Consequently, many of these bridges are now
distressed and repair methods are needed to restore them. This
paper presents the results of a research project, funded by the
Pennsylvania Department of Transportation (PADOT) and the Federal
Highway Administration (FHWA) for the purpose of exploring methods
to restore deteriorated bridges composed of prestressed, concrete,
adjacent box beams.

Background of Problem

Over 600 adjacent box beam bridges have been built in

Pennsylvania. Many were constructed before 1955. A typical
bridge has nine beams, each 3 feet wide and may have several
simply supported spans. Mortar filled shear keys and lateral tie
rods help to transfer concentrated vertical loads to adjacent
beams.

Two fabrication practices used before 1955 have led to the present
state of deterioration of this bridge type. First, calcium
chloride was often added to the concrete mix to accelerate curing
in cold weather. Calcium chloride, in the presence of air and
water, causes corrosion of steel items embedded in concrete. The
corrosion causes pitting of the prestressing strand and loss of
cross-sectional area. The corrosion product, rust, exerts
bursting forces on the surrounding concrete leading to cracking
and spalling.

Second, forming practices were uneven. The prestressing strands

were not provided with sufficient cover, particularly on the beam
soffits. Adequate cover of high quality concrete will normally
protect the embedded steel from corroding by blocking the passage
of water. Water is able to reach the beam soffits because of
leakage through joints between the box units.

An extensive literature search (References 1-13) was made to

discover previously tried repair methods. Nothing applicable to
adjacent box beam bridges was found.

Two methods of repair were developed and installed on a test

bridge near York, Pennsylvania. One method used deformed
reinforcing bars and the other, which is the subject of this
paper, used post- tensioned strands. The repaired bridge was
tested to determine the overall behavior of the bridge and to test
the efficacy of the repairs.
 Prestressing 231

DESCRIPTION OF TEST BRIDGE

The test bridge was chosen by PADOT and was considered to be

representative of this bridge type. The bridge carries Locust
Point Road (L.R. 66038) over the Little Conewago Creek in York
County, Pennsylvania. The bridge consists of three approximately
52 foot spans and is skewed 30 degrees. Each span has nine
adjacent box beams, each 36 inch wide and 33 inch deep. A 24.75
inch diameter circular void was at the center of each box. Edge
beams have 10 inch by 10 inch curbs and steel guard rails. An
asphalt wearing course provides the driving surface. Buried
expansion joints are used over piers and abutments. Staggered tie
rods join the boxes at span third points. Figure 1 shows plan and
cross-sectional views of the test bridge.

Bridge bottom reinforcement, according to state drawings,

consisted of two rows of 1/4-inch diameter strands, with 40
strands in each row. Top reinforcement consisted of four #6 bars.
Welded wire fabric was used for shear and temperature
reinforcement. Reinforcement details, as designed, are shown in
Figure 2.

The condition of the bridge was evaluated by visual inspection,

optical survey and core inspection and testing. Many spalls were
observed on beams soffits. Within the spalls, corroded strand was
seen. Often the spalls extended for some considerable length.
In these cases many strands were corroded through and were hanging
down from the soffit. In the most severely deteriorated beam,
less than 20 strands were exposed. Most spalls were quite
shallow, generally less than .4 inches. The deepest spalls, up
to about . 8 inches deep, were located adjacent to joints.
Numerous damp areas were also seen, indicating the leakage of
water through joints. Gaps at joints ranged from .1 to .8 inches
wide. On the top surface of the bridge, longitudinal cracks
occurred above joints between box beams and over buried expansion
joints.

Of the three spans, the middle span was considered to be in the

worst condition. However, test repairs were made to the south span
since this was the only span completely above dry land. A
reflected soffit plan showing the observed condition of the south
span is presented in Figure 3.

Cores were taken at a joint and at an end diaphragm. The joint

core showed that mortar in shear keys was of poor quality. The
top of the key was filled with asphalt. The bottom of the key was
filled with gravel. The keys had an unusual profile as shown in
Figure 2. The core also showed that the asphalt wear course
consisted of many layers with a total thickness of 5 inches.
Beneath the top layer, the asphalt was fragmented.
232 Osborn and Preston

The diaphragm core (see Figure 3 for location) was for the full
depth of the member. Chloride ion contents and concrete
compressive strengths were measured from the core. Chloride ion
contents ranged from 6. 5 lbs to 11.2 lbs per cubic yard of
concrete. These levels are 5 to 8 times those needed to support
corrosion of embedded steel. Compressive strengths measured on
two samples were 7700 psi and 8900 psi. Based on the compressive
strengths, it is estimated that the flexural tensile strength is
over 800 psi and the modulus of elasticity is about 4,700,000 psi.
Concrete of this strength is relatively impermeable. Two layers
of strands were found at the bottom of the core. No corrosion was
visible.

An optical survey of the soffit of the south span showed the

general bridge profile. The bridge had a longitudinal slope of
about 1.3 percent, a cross slope of .7 percent and a camber of
about .05 inches.

A sample of corroded strand was taken for tension testing. It

broke at a tensile load of 5800 lbs. The original strength of the
strand was probably about 9000 lbs. The area of the strand at the
break was about . 023 square inches compared to the original
diameter of about .035 square inches. Thus the strength of the
corroded strand was consistent with its reduced cross-sectional
area. However, the corroded strand was quite brittle with an
ultimate elongation of only .65 percent and a reduction of area
of 0 percent. The original strand, which was not stress-relieved
at the time these bridges were built, could be expected to have
an elongation of 2. 0 percent and a reduction of area of 30
percent.

REPAIRS TO TEST BRIDGE

The stated objective of the project called for the development of

several repair concepts which would restore the original ultimate
strength of the bridge beams. Although having merit, approaches
which did not restore ultimate strength of the individual beams
were not to be considered. This condition precluded the use of
concepts which employed new analytical tools, relaxed code
requirements or repairs to redistribute moments and shears.

Two repair methods were chosen for implementation. The selection

process was based on viability and cost. Project funds were only
sufficient for the repair of two beams from one span. Each beam
was to receive one of the chosen repair methods.
 Prestressing 233

Alternate Repair Concepts

Four repair concepts were developed. Two used external post-

tensioned reinforcement and two used external non-post-tensioned
reinforcement. The four approaches are listed below.

External Post-tensioning Approaches:

1. Four .5 inch diameter straight strands, tensioned against

end anchorages bolted to beam soffits, then gunited after
tensioning for composite action.

2. Two .6 inch diameter strands draped from the top, through

beam joints, which followed the soffit in the middle third
of the beam. Also gunited after tensioning.

External Non-post-tensioning Approaches:

3. Six #6 reinforcing bars, suspended below the soffit and

gunited to provide composite action.

4. One 18 inch wide by .125 inch thick steel plate, bonded to

the soffit with epoxy and further anchored with bolts.

Approaches 2 and 4 were rejected primarily because of anticipated

construction difficulties with associated higher cost. References
17 and 18 contain more complete explanations of the selection
process. This paper discusses the installation and testing of
approach 1. Installation and testing of approach 3 was similar
and is described in detail in Reference 18.

Design of Post-tensioned Repair

Original design calculations were not available. It is likely

that the original design was based on AASH0-1949 Specifications
and "Specifications for Prestressed, Precast Reinforced Bridge
Deck " dated May 1, 1952, by PADOT. Design parameters for a
similar bridge built at the same time were listed in Reference 2,
as follows:

o Design loads per beam = 80 percent of the wheel load

from an HS20-Sl6-44 truck plus impact

0 Allowable tensile stress 0 psi

o Area of 1/4 inch strands = 0.0352 square inches each

o Initial strand tension= 135,000 psi

o Strand tension after losses was assumed to be 80 percent of

initial tension
234 Osborn and Preston

These parameters established the strength requirements for the

repaired beam. A prototype deteriorated beam was assumed to have
the following deficiencies:

o Bottom row of 40 strands broken

o Bottom 1.25 inch of concrete spalled off

o Top row of uncorroded strands fully effective and located

3-1/4 inches above original bottom

Based on the test bridge, these assumptions of deterioration are

very conservative.

The difference between the original design strength and the

assumed deteriorated strength dictates the additional strength
which must be provided by the repair. The repair design was based
on the addition of .5 inch diameter strands positioned 2 inches
below the bottom.

The final design of the post- tensioned repair called for the
placement of four .5 inch diameter strands, each tensioned to a
load of 18 kips. A lesser number of strands would have been
possible, except that a symmetric pattern would have been
difficult to place. A simple welded plate assembly (Figures 4 and
5) was designed as an end anchorage for the strands. Each
assembly was to be bolted to the beam soffit with six Hilti HAS
Super 78-10 polymer grouted anchor bolts. Uncertainties in the
shear strength of these embedments dictated a conservative design.
Also, the anchorage design was predicated on preventing the
anchorage from being the weak link in the system. A schematic of
the anchorage is presented in Figure 4.

Flexural capacity and midspan bottom fiber stresses for the

original beam, deteriorated beam and repaired beam are listed in
Table 1. Also listed are design requirements according to 1949
AASHO and 1983 AASHTO specifications. The "assumed" conditions
are based on the loss of the bottom row of strands and their cover
as mentioned previously. The "actual" conditions are based on the
average condition observed on the south span of the test bridge
which indicated an average loss of less than 10 strands per beam.

TABLE 1 - FLEXURAL MOMENTS AND BOTTOM FIBER STRESSES

Midspan Bottom
Ultimate Flexural Stress under Dead
Capacity at Midspan and Live Loads
(foot kips) io.ill*
Original Beam
(As designed) 1444 -82**
 Prestressing 235

Table 1 - Flexural Moments and Bottom Fiber Stresses

(Continued)

Midspan Bottom
Ultimate Flexural Stress under Dead
Capacity at Midspan and Live Loads
(foot kips) ..lo.ill*
Deteriorated Beam
(Assumed) 757 -763**

Repaired Beam
(Assumed) 1122 - 307**

Deteriorated Beam
(Actual) 1295 +24***

Repaired Beam
(Actual) 1718 +266***

Code Requirements
Required Ultimate Required Midspan
Flexural Strength Bottom Fiber Stress
(foot kips) ..lo.ill
AASHT0-1949
Specification 1056 0

AASHT0-1983
Specification 777 -424

* + - compressive stress
- - tensile stress
stress calculation based on AASHT0-1949 truck load
distribution
stress calculation based on measured truck load distribution
simular to AASHT0-1983

The durability of repairs was an important design consideration.

For this reason all exposed steel hardware was epoxy coated. The
gunite is expected to protect uncoated steel hardware. The gunite
also eliminates the hazard from a strand should it break in
service. On this project, this hazard is minimal since there is
no traffic under the bridge.
236 Osborn and Preston

Construction of Repair

The Post-tensioned repair was made in 9 stages as follows:

1. Loose and delaminated concrete was removed and loose

prestressing strands were cut off.

2. The beam soffit was sandblasted to clean off contaminants

and improve bond of the gunite. Exposed strands were
sandblasted to near white metal.

3. "L" shaped anchors were set into drilled holes. These were
used to hold up the mesh reinforcement and help provide
composite action between the original beam and the gunite.

4. The post-tensioning anchorage frames were installed. A

problem drilling the bolt holes for the frame was
encountered during this stage. The frames had been
fabricated and epoxy coated in a factory.

5. Four high bond epoxy coated prestressing strands were

threaded through the frames and tensioned to about 18 kips
each. The ends of the strand were cut off and encased in
sealant filled caps. Bolt heads were painted.

6. Galvanized welded wire mesh was tied to the "L" anchors.

7. Form boards were positioned on either side of the beam

soffit.

8. The 3-1/2 inch thick gunite was sprayed on in 3 layers using

the dry process technique.

9. The soffit of the gunite was screeded and troweled. A

curing compound was sprayed on the gunite.

Features of post-tensioning system

Strands used were .6 inch diameter since wedge anchors for epoxy
coated, .5 inch diameter strands were not available. The post-
tensioning loads, however, remained the same as designated for .5
inch strands.

Sufficient clearance must be maintained between the anchorage and

the pier wall to accommodate the ram. The Dywidag system used on
this project required 36 inches.

An experienced supervisor was retained for the post-tensioning

operation.

The wedges were set and slack was removed by first tensioning each
strand to 1.8 kips. Then each strand was tensioned to 18 kips.
 Prestressing 237

Load in the strand was measured by the calibrated pressure gage

in the hydraulic line and verified by the elongation of the
strand. The strand elongated about 2 inches over its 42 foot
length under full tension of 18 kips. The measured elongation was
used after release of the jack to determine the final force in the
strand. A photograph of the strand tensioning in progress is
shown in Figure 5.

Problems Encountered During Construction

The repairs for a single beam took about one week. They would
have been done sooner except that the contractor had much
difficulty in drilling holes for the anchor frame bolts. The
holes were wider than the spacing between prestressing strands in
the original beam. It was found that the percussion type masonry
bits used by the contractor were inappropriate. The strands were
very hard, which caused the drill bit to become dull. A water
lubricated diamond core bit would have been superior, if one of
the proper diameter was available.

Two of the anchor bolts were found to be loose after installation.

The anchorage strength was downgraded accordingly. If this
problem was more common, required repairs would have been
difficult to install. This illustrated the need for quality
control for critical items. A revised design which allowed for
a percentage of unusable anchor bolts would be prudent.

Another potential problem was in the use of the "L" anchors. The
standard form of these anchors have lead expansion bushings. The
combination of lead, steel, zinc coating and dissimilar concretes
can create a galvanic cell which will corrode the anchors. Other
inert anchor materials should be considered for future projects.

A crack developed between the gunite and the original beam soffit.
However, it did not appear to affect the composite behavior of the
repair. Before attempting an actual construction project, other
methods of applying the gunite should be explored. Cores should
be taken through the finished work to verify proper bond.

Cost of Repair

Since this was a research project and involved a relatively small

area, construction unit costs were high. The cost to the project
amounted to about $100. per square foot of soffit surface (1985).

The contractor provided an estimate of $78. per square foot to do

a larger area. This price also seems high and may reflect the
contractor's difficulty in installing the post-tensioned repair.
238 Osborn and Preston

LOAD TESTS ON BRIDGE

The primary purpose of the tests was to verify that the gunite
encased external reinforcement was acting compositely with the
original beam under service loads. The tests were also designed
to yield information about the lateral distribution of loads
across the bridge, the distribution of strain through the cross-
section of a beam, and the distribution of strain along the length
of a beam.

Instrumentation and Data Acquisition

Devices used: Instrumentation devices consisted of 2 inch long

bondable strain gages and 1 inch (span) x .001 inch (resolution)
dial gages. These devices were concentrated on repaired beams 3
and 7, as shown on Figure 6. They were also distributed to the
other beams so that useful information about lateral distribution
could be obtained. The instrumentation layout is presented in
Figure 6.

Strain gages were put on top surfaces of Beams 1 and 3 to measure

strain distribution through their cross section. Four gages were
placed near the end of Beam 5 to pick up negative moment effects.
To simplify measurement of strains at the level of the epoxy-
coated strands, short reinforcing bars called sister bars, with
strain gages attached, were embedded in the gunite.

Strain gages were monitored with a Vishay Model P3500 Strain

Indicator coupled with two Model SB-10 Switch and Balance units.

Installation: Weights suspended from the soffits by Invar wire

provided a stable reference for the dial gages. Dial gage stands
were mounted on steel plates bonded to concrete blocks which were
partially buried in the soil.

The soffits were prepared for strain gages by grinding, applying

a layer of epoxy and sanding. The epoxy layer provided a smooth,
continuous bonding surface for the gages. All gages were
waterproofed for long term stability so that the test could be
repeated in later years, if desired. Three-wire cables connected
the gages to the data acquisition system. Both dial gage and
strain gage readings were recorded manually.

Tests Conducted and Magnitude of Applied Loads

Tests consisted of 1) a pre-repair load test using a single ten-

wheel dump truck, 2) measurement of strain and deflection due to
the application of post-tensioning, 3) a post-repair load test
using one and two ten-wheel dump trucks, and 4) a post-repair load
test using one and two twelve-wheel dump trucks. The ten- and
 Prestressing 239

twelve-wheel truck patterns and axle loads are shown in Figure 7.

The ten-wheel trucks weighed about SO kips. Two such trucks, back
to back, simulate the effects of HS-20-44 loading plus impact.
Each twelve-wheel truck weighed about 7S kips and simulated a ML-
80 load without impact. All tests were conducted between July 18,
1986 and October 23, 1986.

Application of Load and Measurement of Response

External post-tensioning forces were applied in 4 stages for each

of the four strands. Strain and deflection readings were taken
at each load stage.

Vertical load was applied by centering the truck wheels over

individual beams across the bridge at midspan. Patterns for ten-
wheel trucks centered over an edge beam and over the middle beam
are shown in Figure 8. All loads were applied with the trucks
stationary to simulate static conditions.

The strain and dial gage readings were taken with the truck(s) in
each position and under zero live load, between each position,
with the truck(s) off the bridge. Strains were converted to
stress by multiplying by the estimated modulus of elasticity,
4, 700,000 psi.

Measurement of Response from Post-tensioning

Figures 9a and 9b show the strains and deflections measured at the

midspan soffits after application of all post-tensioning forces.
Figure 9c shows the relative distribution of strains and
deflections to each beam. Total force, after seating of the
strands and immediate losses, was about 70 kips. This force
created about 20 microstrain at the bottom of the repaired beam
and caused an upward deflection of about . OS inches. Due to
transfer of loads laterally to adjacent beams, only about 2S
percent of the force was resisted by the repaired beam. In an
actual repair of the entire structure, the relative distribution
to each beam would be practically uniform.

Pre-Repair and Post-Repair Test Results

Results are presented for just two load positions: one with the
truck(s) along an edge beam and one with the truck(s) centered on
the bridge. Results for the remaining positions can be found in
Reference 18.

Demonstration of Composite Behavior: The post repair tests

revealed the composite behavior between the original beam and the
gunite encased repair. Figure 10 shows the distribution of strain
240 Osborn and Preston

through the cross section of the repaired beam under single and
double truck loading. The strain distribution is nearly linear
which indicates that the reinforcement within the gunite is acting
compositely.

Lateral Load Distribution, Ten-Wheel Dump Truck Loadings (HS 20-

44): The pre-repair and post-repair test results were essentially
the same with respect to lateral distribution of loads. Figures
lla and llb show deflection and strain across the bridge with the
two trucks positioned as close to the edge beam as possible. The
highest deflection is .14 inches and the highest strain is 70
microstrain for an estimated stress of about 330 psi. Both occur
in the edge beam under load. Figure llc shows that the edge beam
carries about 33 percent of the total truck load. (This is
equivilent to 66 percent of the total wheel load on one side of
the truck. ) With the trucks in the middle of the bridge, the
deflections and strains in the beam under load are much less, as
shown in Figures 12a and 12b. The distribution is much wider and
the loaded beam is only resisting about 30 percent of the truck
wheel load.

Twelve-wheel Dump Truck Loadings (HL-80): Deflections and strains

are shown in Figures 13a and 13b which reflect the loading of
standard AASHTO lane positions, with the lanes as close to one
edge as allowed. Figure 13c shows that the edge beam is only
carrying about 14 percent of the weight of both trucks. Since
there are four sets of wheels in the two trucks, this beam resists
56 percent of a wheel load for this load pattern. This is the
highest measured percentage for any combination of AASHTO load
patterns.

Measurements of stresses along the length of Beam 5 are presented

in Figure 14. The negative bending of the beam near its support
is indicated by the stress reversal in the figure.

Visual Inspection of Structure During Load Test

No new cracking of the original beams was observed during the

test. This is not surprising since analysis indicates that dead
and live load stresses plus the original prestress are
compressive. In the repaired beam additional compressive stresses
are generated after post- tensioning. However, the guni te is
unstressed by the post-tensioning. Some transverse cracking did
appear in the guni te. This is attributed to shrinkage and
flexural tension.
 Prestressing 241

COMPARISON OF TEST RESULTS WITH ANALYTICAL STUDIES

Stresses and deflections were computed for simple span beam

members of the type used in the test bridge. Computed values were
about 20 percent higher than measured values. This differential
is higher than the uncertainty in the modulus of elasticity. A
more rigorous analysis was conducted using the "Transfer Matrix
Method" described in Reference 16. This analysis showed that
negative moments are developed at beam ends due to the skewed
design of the bridge. The magnitude of negative moments is
influenced by the position of load. This finding is consistent
with the measured behavior of beam 5. The lateral distribution
of loads to beams was also accurately predicted by the Transfer
Matrix Method.

DISCUSSION AND CONCLUSIONS

The external post-tensioning of these box beams was shown to be

a viable method of repair. Load tests conducted after repair
showed that gunite can be used to develop composite interaction
between the original beam and the external strands under design
load levels. This does not prove that the composite action can
be assured under ultimate load conditions. The long term bond of
the gunite to the beam soffit has also not been established.
Additional studies are needed to assure that composite action is
maintained both under ultimate load conditions and over the long
term.

The problems encountered in drilling through prestressing strand

point to the need for different methods of attaching post-
tensioning anchorages. It is possible that a larger number of
smaller diameter anchors could have been used to avoid these
problems. Also, there may be drill bits, unknown to the authors,
which can more easily penetrate the existing strands. In
addition, costs would be lower if only two anchorages per beam
were needed instead of four. Research is needed to establish the
shear strength of multiple bolts embedded in concrete where the
bolts are aligned with the load path, where the bolts are embedded
into a relatively thin bottom flange and where the bolts are
positioned close to a transverse edge.

One of the most enlightening results of the test program was the
measured lateral distribution of loads to beams adjacent to the
loaded beam. In spite of the poor condition of the shear keys,
lateral distribution was nearly equal to the distribution
designated in current AASHTO specifications and that determined
from analysis. This means that even bridges which appear to be
severely distressed may have substantial reserve capacity since
they were designed using conservative assumptions with respect to
lateral load distribution and allowable tensile stresses. The
242 Osborn and Preston

original design of the test bridge was found to exceed current

design requirements by 57 percent as a result of these factors.
In spite of the apparent poor condition of the bridge, a complete
evaluation revealed only a 15 percent reduction in its original
strength.

ACKNOWLEDGEMENTS

The authors wish to thank PADOT for their support of the research
project which is the basis of this paper. Many individuals at
Wiss, Janney, Elstner Associates participated in the research.
However, the authors wish particularly to thank Carol Roach for
her substantial efforts on this project.

REFERENCES

1. AASHTO Standard Specifications for Highway Bridges, Thirteenth

Edition, 1983. Published by American Association of State
Highway and Transportation Officials, Washington, D.C.

2. Roesli, A.; A. Smislova; E.G. Ekberg, Sr.; and W.J. Eney, all
of Lehigh University, Field Tests on Prestressed Concrete
Mul tibeam Bridge, Highway Research Board Proceedings, Vol. 36,
1956.

3. Zollman, Charles C., Dynamic American Engineers Sustain

Magnel's Momentum, PCI Journal V. 23 No. 4, July-August 1978,
pp 47-51.

4. Shanafelt, G.O. and Horn, W.B., Guidelines for Evaluationn and

Repair of Prestressed Concrete Bridge Members, National
Cooperative Highway Research Program Report 280.

5. Shanafelt, G.O. and Horn, W.B., Damage Evaluation and Repair

Methods for Prestressed Concrete Members, National Cooperative
Highway Research Program Report 226.

6. Ladner, M. , Concrete Structures with Bonded External

Reinforcement: Report No. 206. M. Ladner/Ch. Weder, Massive
Construction Section EMPA, CH-8600 Dubendorf, Switzerland.

7. Dussek, I. S. , Strengthening of Bridge Beams and Similar

Structures by Means of Epoxy-Resin-Bonded External
Reinforcement, Transportation Research Record 785.
Transportation Research Board.

8. Irwin, C.A.K., The Strengthening of Concrete Beams by Bonded

Steel Plates, Transport and Road Research Laboratory,
Department of the Environment, Old Workingham Road, Crowthorne
RGll 6AU Berkshire, England.
 Prestressing 243

9. L'Hermite, R., Use of Bonding Techniques for Reinforcing

Concrete Structures, Conseiller Scientifiguede 1 U.T.I.
(Union Technique Inter-Professionnelle du Batiment et des
Travaux Publics) Ancien Directeur de la Rechevche. In French
with English synopsis.

10. Koretzky, H.P., What Has Been Learned from the First
Prestressed Concrete Bridges Repair of such Bridges.
Transportation Research Record 664, Bridge Engineering, V. 1,
Transportation Research Board.

11. Rabbat, B.G., Kaar, P.H., Russell, H.G. and Bruce, R.N., Jr.,
Fatigue Tests of Pretensioned Girders with Blanketed and
Draped Strands. PCI Journal, V. 24 No. 4, July-August 1979.

12. Overman, T.R., Breen, J.E. and Frank, K.H., Fatigue Behavior
of Pre tensioned Concrete Girders, Research Report 300-2F.
Center for Transportation Research, University of Texas at
Austin No. 1984.

13. Irwin, C.A.K., Static and Repetitive Loading Tests on Full

Scale Prestressed Concrete Bridge Beams, Report No. 802.
Transport and Road Research Laboratory, England 1977.

14. Sanders, Jr., W.W., and Elleby, H.A.,. Distribution of Wheel

Loads on Highway Bridges, National Cooperative Highway
Research Program Report 83, 1970.

15. ACI Committee 363, State-of-the-Art Report on High-Strength

Concrete. J. of the American Concrete Inst., V 81 No. 4,
July-August 1984.

16. Buckle, I.G. and Lee, Weng-Onn, Analysis of Skew Multibeam

Bridges by the Transfer Matrix Method, International
Conference on Short and Medium Span Bridges, Toronto, 1982,
Volume 2. Published by Canadian Society for Civil
Engineering, Montreal, P.Q., Canada.

17. Preston, H.K. and Osborn, A.E.N., Restoration of Strength in

Adjacent Prestressed Concrete Box Beams, Interim Report No.
FHWA-PA-85-016 to the Pennsylvania Department of
Transportation, Wiss, Janney, Elstner Associates, Inc.,
January 31, 1986, 50 pp.

18. Preston, H.K., A.E.N. Osborn and C.E. Roach, Restoration of

Strength in Adjacent Prestressed Concrete Box Beams, Final
Report No. FHWA-PA-86-044+84-21 to the Pennsylvania Department
of Transportation, Wiss, Janney, Elstner Associates, Inc., May
14, 1987, 122 pp.
244 Osborn and Preston

BE.AM
NUMII[RS I0!0Jc5J@j0j0!0!0!&TI::~:
I e" 3 • -o" • 27' -o" 1 DESIGN WIOTH l ' .J
I NOTE· MEASURED WIDTH

v
BRIDGE CROSS· SECTION WAS 27'-4" DUE TO
GAPS BETWEEN SOX
(LOOKING NORTH) BEAMS

~ Iff d /TIE BOLTS

~
.. iff
..
....
" •'
·'

•'
•' •'

"
.. ,.
·• ..
•'
' ct>
~ ~
51'-o" o/c IRGS A 51'-o" 51'-o"
l~~
52'• 5" 52'-9" 52'- 5"

BRIDGE PLAN

Fig. 1 - Plan and cross section views of test bridge.

 Prestressing 245

1' s"

2/5-:3"x 8" WIRE MESH IN

END QUARTERS OF BEAM
4/5·:3"xe" WIRE MESH IN
MIDDLE IOF BEAM

4- •s BARS FULL
I LENGTH 1/2'' ( AVG. MEAS.)

• ..... ......

..

•• (I)
,.,~
,o ..
'V
.. '<t
....
"'<t

-.- .l.::-:-:-:-:-:-
~
.................... ····· ............ . ..
........................ . ~:.

-~
-
l 40 -1/4"2) STRAND PER ROW
APPROX 314" C. TO C.
I JL "
112'' AVERAGE
MEASURED GAP

Outside dimensions are based on actual measurements.

Reinforcement and prestressing strand layouts are
based on standard drawings.
Fig. 2 - Typical box beam.
 t-,j
.., ~
NaTe : @ JOINT ~ 4"" OPCt.l I"@ MID n.1R.D
'" """'
"' 0'1
® t.e4M 7 +'+" DOWN ~L.A.TIVC.
TO cc.AM (Q @. .JOINT
~OUTH :>Pto.~ U>JDr!R~IDr!
0rJl
~o·- 4'1
't. CI(.F'.•rr (tiP) ti'
0
a
ll:l
::l
Q..
1'""4.Ceo.-
P•<=
~
('b
rJl
.....
0
::l

I,.
<B:> vv..•
eli:> 1/'!&"
ul

LC.R .oo5 =>N

--~~~------~~-=R·~~
0

.
--- '
ec.o.Le
• •

(~T)

@+/¥o" - COAe.te> T.,Ke ....

@.,jv,:·

· CuiOleA~

~ eo- ~D :>~o- ..,.a~ .-rru.No~jcoveR - . .IANNlY, a.ana ..aaATU, aec.

..-..:&,. ece - l:ll(proeeo c01ta0oeo efT'R6.ND
t~- INCtP18NT SP6.1...1...
A6UT..........-
LC.A.- L.ONCI; CJU:tr,.CK_ - \VIOTH NQ'T"eD
L.d - LIMe. DePO'OITO a'Z.-z.- sa LITTLe CONCW'AC:.O CIZC:C:.II( ~IDGC.
hc-~"(COMO LR.~b0~6 '='TA. I.,+&Z. YORK.CO.,PA
TC- ~~ c.a.ctc.
X~- l._.Of>JTATION IN CONCR.C.Te ,.liQM !""'RM
d!!J)- W.&.TeR &POT"e> ·~· en""'o'vN PI:"NN. D~P'T OF" TRAN~PORT6.T\ON.I 6-41Z07

Fig. 3 - Condition survey of bridge

soffit - south span.
 Prestressing 247

',(

~~t~~~====!~~!:~t!J,1z"auNNJrE
WWF2 • xz•
"12fiAI.kiE

a. ~EIIH CROSS·SECTION
SHOWIN6 POST· TENSION£0 REPIIIRS

b ONE ENO ANCHORAGE.

Fig. 4 - Post-tension repair details.

Fig. 5 - Jacking unit tensioning Strand No. 2 at end of

Beam Ro. 3.
 l'-.:)
~
KEY FOR STRAIN AND DIAL GAGES 00
N l l N NOTE: ALL GAGES AT i Of SPAN ALONG SKEW

lll
L DISTANCE FROM FACE Of PIER IN FEET AND i OF lOX BEAM UNLESS NOTED OTHERWISE
~POSITION RELATIVE TO GAGE AT CENTER Of BEAM
£-EAST, W- WEST, N-NORTH (NEAR PIER) 0CfJ
-· - - S·SOFFIT, B-REINFORCING BAR, G-GUNNITE, T-TOP
---- - B£ AM NUMB£ R
0"'
0

BOX BEAM 3'-o"

a
~
SHEAR KEY CURB ::I
0...
~
~
-~
CfJ
.N ......
0
::I

IHVAR WIRE

~~··· ...PlATE
STEEl
. ... ~
\:
;: ~
: o
~=!=!~
: : o 1
A
' I
~
::
r==1
I o :
~
F3: : : ~%~
n
~ I ~~ :===1 ~
o o

~~~EKD c~-'
: : : :
:.. .. 1
! :: :
~,._ __.: !.... •• J
:
'-··J
! :
L-J
: : :
l _ _ _,
! : !L.~:
._ __ J

Beam 3: Pos~-tensioned repair.

Bea. 7: Reinforcing bars repair.
Fig. 6 - Layout of instrumentation on south span
of test bridge, looking north.
 Prestressing 249

13K
t 't
18K 18K

D l,,.tc ___g-

~
c ----'!.
5'1" ~ Je'o"
•1=2,. c.
10"

a. 10-wheel dump truck c. 12-wheel dump truck

used in Test 1 used in Test 3
(one-half HS20-44 (ML-80 loading).
loading).

11 1 11 1 11
1
13 -4 4'-3" e'-3 1/2" 4 -3 13 -4
t• I 'I I I
19K 19K
·I
13 K
13 K 20K 20K

c 0
[-~ ~
D c c D
&-o ~ 10" 7"

a __g__
c
c
c
.rt1t
c ~~1!1' 11t0
-t1""

b. Two 10-wheel dump trucks used

in Test 2 (HS20-44 loading).

Fig. 7 - Wheel patterns and loads for trucks

used in Tests 1, 2 and 3.
250 Osborn and Preston

PATTERN
IW
4W

.. '-

Fig. 8 - Wheel pattern for 10 truck positions, lW and 4W,

for second part of Test 3.
 Prestressing 251

0.15
0 0 0 CD 0 80 ® ®
....;
0.12

.......c"
0.01

u
......." 0.01

~
"
0.05
0.04

0.03
0.02
0.01
....... __./ n ---
........_
0.00

(a)

...____
~I v
/ r--..
-
~

(b)

...
...c01 ..."" .
.....
c orn.tcnoN
.," STRAIN

~
....c 1-c ,..
" 8
..." ......u
00

"u" ........."
1-

"'" &

(c)

Fig. 9 - Measured deflection and strain due

to post-tensioning of Beam No. 3.
252 Osborn and Preston

TWO TRUCKS
ONE TRUCK

TOP OF BEAMJ

BEAM SOFFIT ---.__

-------=--=----=-=::.::-; -----------

INSTRUMENTE0--=:7 -6() -40 -2o o 20 40 60

SISTER BAR
IN GUNITE MICRO STRAIN

Fig. 10 - Distribution of strain through

the cross section at Beam No. 3.
 Prestressing 253

ill] ill (

0 0 0 0 00 0 0 0
0.00
-0.01
-O.Ol
v - f.--.>

~r
....~ -o.oJ
-0.0-4

...."
..0
-o.oe
u -0.07

......." -o.oa
..
v ~
-o.oe
= -0.10

-0.11

-0.12
-0.1.l

-0.1S
I
Fig. lla

II
\ I
I
.. \ ~

,. 1\_1--.-
t---.-_
i
~
r--
Fig. llb

I
I

... ....".. <OX~ .

c DEFl...ECTION
STR/.11>1

.. ...
I'll
0
1-<
.. I
l
....0 "'...0
~011

....... ........
.. g 20llli ~
I

u
"... .......
u
.. ..
"' =
Fig.

Fig. 11 - Distribution of deflection and strain for

two truck load patterns lW in Test 2.
254 Osborn and Preston

:\ m- m ;

0.00
0 0 0 000 0 00
-0.01

....,;
-0.02
-o.u ....... /
v v
c
-0.0<1
-o.oa
~ ...............
..
.,..0 -0.01

~
.........
u -0.07

-o.oa
-0.01
"
1'1 -0.10
-0.11
-0.12
-G. I>
-0.14
-G.IO

Fig. 12a

••

..••
.,..c
.."'.... ..
..
0 .. II r--- "'- 1--
.. -
....)1:u J--.... !'---
20
r--.--..

Fig. 12b

...... .,.." ... I I

.. .....
0 0£"-ECnON
-+ STJWN

...'"' "'..
..
0 0

....:0 .,....g ,
".. ......."
u
u

... -; v-- t--.-..... t:::-.

~ I
""" 1'"1 <>-')..--
~
:
Fig. 12c

Fig. 12 - Distribution of deflection and strain for

two truck load patterns 4W in Test 2.
 Prestressing 255

....,; -o.o•
-o.o2
-o...
......."
0 -o...
u -o.oe
....
~

....
-o.OI
-o.07
~ -o.oo
-o.o•
-o.10
-G. II
...0.12
-0.1~

-o.,.
-o.u

..
..
..
...."
...... ..
"...0
....u ..
lC
.
..

Fig. 13 - Predicted deflection and strain due to

two lanes of HS20-44 loading lanes
placed on the west side of the bridge.
256 Osborn and Preston

80

10

60
face of
pier ~ of bea• 2_
50

40

30 -

I
20
"
I'" 10

~~

"'
'''
·-20 ----,--r- -.---~-~-~--,----,--·..--,.--·1---,--

o 4 8 12 16 20 24

POSITION ALONG BEAM (FT)

Fig. 14 - Strain measured on soffit of Beam Mo. 5

due to load pattern 5W.
 SP 120-12

Deviator Behavior and Design for

Externally Post-Tensioned Bridges
by R.J. Beaupre, L.C. Powell, J.E. Breen,
and M.E. Kreger

Synopsis: A laboratory investigation was performed to study deviation saddles, a type

of tendon deviator used in externally post-tensioned precast segmental box girder bridges.
Ten reduced-scale models of deviation saddles were fabricated and loaded to ultimate using
a specially designed testing apparatus that applied load to each deviator just as would be
applied to a deviator in a bridge. The objectives of the study were to:
( 1) experimentally investigate the strength and ductility of deviators;
(2) evaluate deviator details in light of observed performance;
(3) define behavioral models for deviators;
(4) determine the effects of using epoxy- coated reinforcement; and
(5) establish design criteria. Data from the test series are presented, two analysis
techniques are formulated and design recommendations are made for design of
tendon deviators.

Kevwords: box beams; bridges (structures); mcxiels; I1Q§t::

tensioning; precast concrete; prestressing steels; segmental
constru.ction: structural design; tests; unbonded prestressing

257
258 Beaupre et al

ACI member Richard J. Beaupre is a Design Engineer for Greiner Engineering, Inc.
in Tampa, Florida. He received a BS in Civil Engineering from the University of Florida
in 1984, and a MS in Civil Engineering from The University of Texas at Austin in August
1988.
ACI member Lisa Carter Powell is a Senior Bridge Designer for Figg and Muller
Engineers, Inc. in Austin, Texas. She received a BS degree in Civil Engineering from the
Georgia Institute of Technology in 1984, and a MS degree in Civil Engineering from The
University of Texas at Austin in August 1987.
John E. Breen, FACI, holds the Nasser I. Al-Rashid Chair in Civil Engineering at The
University of Texas at Austin. He is the past Chairman of ACI Committee 318, Stan-
dard Building Code, and is a former chairman of ACI-ASCE Committee 441, Reinforced
Concrete Columns, and of ACI's Technical Activities Committee
ACI member Michael E. Kreger is an Assistant Professor of Civil Engineering at The
University of Texas at Austin. He received BS, MS and Ph.D. (1983) degrees from the
University of Illinois at Urbana-Champaign. He is a member of ACI Committee 318-C,
Analysis, Serviceability and Safety, ACI-ASCE Committee 352, Joints and Connections,
and ACI Committee 215, Fatigue of Concrete. He has conducted research in the areas of
repair and strengthening of concrete structures, behavior of reinforced concrete buildings
subjected to seismic loads, and behavior of prestressed concrete bridge structures.

INTRODUCTION

One of the latest and most dramatic developments in segmental technology has
been the use of external tendons which are defined as tendons in ducts not encased in the
concrete of the webs or flanges of the box girder bridge and attached to the concrete boxes at
only a few discrete points (see Fig. 1). This innovative type of construction has been shown
to provide substantial economic savings, as well as savings in construction time. External
post-tensioning differs from internal post-tensioning because the tendons are removed from
the webs and flanges and placed in the cell-void. The tendon deviators maintain the draped
profile of the external tendons and provide the only positive attachment of the tendons to
the structure other than at the anchorage zones. This makes the deviator a key element of
this bridge system.
Three basic kinds of tendon deviators have been utilized in externally post-
tensioned segmental box girder bridges; the diaphragm (Fig. 2a), the rib or stiffener
(Fig. 2b), and the saddle or block (Fig. 2c). The deviators are usually monolithically
cast in the bridge segments to accommodate the required tendon duct configurations. The
diaphragm and rib or stiffener are usually full web-height deviators. The diaphragms usu-
ally extend the entire width of the box section and are provided with an access opening for
passage along the span. The rib or stiffener extends out only a small distance from the web
wall. The advantage of using th~ diaphragm or rib type deviators is that compressive strut
action in the concrete may be utilized to resist the tendon deviation forces. Compression
struts can develop from immediately above the tendon duct to the top flange which provides
these deviator types with greater inherent resistance than the saddle or block. However,
 Prestressing 259

many disadvantages also exist with these types. They create added dead load, sometimes
offsetting the savings from the efficient web thickness. Other disadvantages are construc-
tion related. The formwork for the diaphragm and rib and the geometry for the tendon
pass-throughs becomes very complicated, especially for a curved span because the bridge is
curving while the tendons remain on straight paths. In contrast, the block or saddle is usu-
ally a relatively small block of concrete located at the intersection of the web and bottom
flange. Advantages of utilizing this type of deviator in a bridge are that there is relatively
insignificant additional weightfoJ:.J.~ structure, the formwork is less complicated than
that r~quired for a diaphragm or a rib, and g~metry complications are minimized because
non-deviated tendon pass-throughs are generally not required. However, the disadvantage
is that the deviator capacity may be greatly reduced as compared to the diaphragm or rib
because there is no major direct compression strut formed after cracking. Therefore, the
deviator force must be tied back into the box by reinforcement. This may require greater
attention to detailing and may lead to more congestion of reinforcing for a saddle-type
deviator than for a diaphragm or rib.

RESEARCH SIGNIFICANCE

Uncertainties exist concerning the behavior and proper design criteria for the ten-
don deviator details. In order to provide careful verification of behavior and develop efficient
details, the study documented herein encompassed an experimental investigation 1 •2 •3 •4 of
the tendon deviator details and suggested a design methodology for the deviators.

DEVIATOR MODELING AND SPECIMEN DESCRIPTION

The laboratory investigation of tendon deviators focused on deviation saddles since

they are the inherently weakest of the three basic types. It was also felt that if the safety
of deviation saddles could be confirmed by this investigation, the wider use of this type
would offer the most advantages through reducing the structure weight, facilitating the
fabrication of segments, and minimizing geometry complications.
The experimental program included fabrication and testing to ultimate of ten
reduced-scale models of deviation saddles. The scale factors utilized were 1/3 for the first
six models and 1/5 for the last four models (reduced since failure forces were approaching
test setup capacity). Two deviator test specimens were fabricated inside a single cell box
section. They were located on opposite sides of the box section at the intersection of
the bottom flange and web wall. The box section was a typical single cell box girder
bridge section except that the cantilevering wings were omitted. The basic box section
design was based on the AASHTO Bridge Specification5 , and remained the same for all
specimens. The box has little influence on the behavior of the immediate deviation saddle
zone. Supplementary box reinforcement was placed in the bottom flange and web at the
deviation saddle for local load distribution.
Deformed microreinforcing bars were used to model the usual #4 and #5 prototype
reinforcing. Microreinforcing bars range in sizes #1.25, #1.5, and #2 which refer to their
nominal diameter expressed in eighths of an inch. Reinforcement test results are presented
in Table 1. Epoxy coated reinforcement which was utilized in several of the test specimens
had a measured coating thickness of 7 mils. A 6000 psi concrete mix with a maximum
aggregate size of 3/8 inch was utilized. The average 28-day compressive strength of the
concrete for each test specimen is presented in Table 2. The rigid tendon duct of the
prototype deviators was modeled with a 1-1/2 inch nominal diameter electrical conduit
260 Beaupre et al

which was bent to required deviation angles using a hydraulic pipe bender. Multiple 3/8-
inch diameter strands were used for tendons in loading the specimens.
Specimens lA and lB modeled a typical prototype deviation saddle detail used in a
straight-span bridge (see Fig. 3). The reinforcement scheme consisted of primary link bars
supplemented with two types of stirrups (designated as open stirrup and closed stirrup).
For specimen lA, three tendons were deviated which represented a tendon configuraton of
a deviation saddle located closest to the center of a span. The corner tendon had both a
vertical deviation and a slight horizontal deviation directed away from the web, while the
other tendon had only horizontal deviation directed towards the web. The specimen lB
reinforcement scheme was identical to that of specimen lA. However, the tendon configu-
ration was different. This deviation saddle only deviated two tendons which represented a
tendon configuration typical of a deviation saddle located somewhat closer to the piers. The
corner tendon had both a vertical deviation and a horizontal deviation directed away from
the web, while the other tendon had only a slight horizontal deviation directed away from
the web. The objective for specimens 2A and 28 was to isolate the behavior of the indi-
vidual reinforcement patterns of specimens lA and lB. These specimens were not intended
to be properly detailed deviation saddles, and they were expected to have an abnormally
low factor of safety at ultimate. Reinforcement details for specimen 2A provided the link
bar alone. Specimen 28 reinforcement details provided the two types of stirrups (open and
closed stirrups) without the link bars. The tendon pattern for specimens 2A and 28 was
identical to specimen lB. The objective of specimens 3A and 38 was to determine if the
frequently used epoxy coated reinforcement has any effect on the behavior and strength of
a deviation saddle. Specimens 3A and 38 were companion specimens to specimens lA and
lB. The only planned difference between these specimens was that the reinforcement was
to be epoxy coated, but some very minor differences in the bar chairs and strain gage lead
wires developed that were not discovered until after fabrication 2 • In spite of this, it was
felt possible to make direct comparisons between the epoxy coated specimens (3A and 38)
and the uncoated specimens (lA and lB).
The objective of specimens 4A and 48 was to evaluate the modified reinforcement
scheme and deviation saddle geometry shown in Fig. 4. This was an attempt to simplify
and standardize reinforcement patterns and deviation saddle geometry for typcal deviation
saddle details. Because of reinforcement congestion, the link bars previously anchored at
the intersection of the web and flange centerlines were replaced by loops anchored into
the expanded nodes and outer stirrups tying the nodes to the web and flange. The actual
reinforcement patterns utilized were based on small rectangular closed loops which enclosed
each tendon, and outer closed stirrups which enclosed the entire deviation saddle. These
bars were anchored under the top mat of reinforcement of the bottom flange. The deviation
saddle geometry was changed to a horizontal top surface with vertical sides. The tendon
configuration for specimen 4A was representative of a deviation saddle on the outside of
a small radius curve. The tendon configuration for specimen 48 was representative of a
deviation saddle on the inside of a small radius curve. Both specimens had two tendons
which had both vertical and horizontal deviations. The objective of specimens 5A and 58
was to further evaluate the effect of epoxy coated reinforcement. Specimens 5A and 58
were companion specimens to specimens 4A and 48 with the only difference being that the
reinforcement was epoxy coated.
 Prestressing 261

TESTING PROCEDURE

A specially designed testing apparatus shown in Fig. 5 applied load to the deviator
just as it would be loaded in a bridge. This load was applied incrementally. The generalized
test setup could accommodate a variety of specimen si:.;es, tendon layouts and loading
schemes. Specimens were usually loaded in two cycles. The first load cycle generally
continued until yield of the reinforcement, and the second load cycle continued to failure
of the specimen. Strain gages were placed internally on the reinforcement of the deviation
saddle to determine contributions of individual reinforcement bars.

TEST RESULTS

Physical behavior of each specimen was observed and noted for the full range of
loadin@J. This included general observation of the deviation saddle, their cracking pattern,
reinforcement fracture locations, and strain data. Detailed results for each test are provided
in References 2 and 4. Typical test results and photographs before and after failure are
shown in Figs. 6-11.
The symbol D and e respectively correspond to the magnitude and direction of
the deviation force on the deviator. The positive horizontal axis is directed towards the
center of the box, and the vector direction is measured clockwise from this axis. The symbol
Do is the nominal design jacketing force for the specimen (0.8~uAp,) based on the total
allowable force of the prototype tendons. The symbol e. is the nominal design jacking
load vector force direction. Since external tendons are basically an unbonded type of post-
tensioned construction, it is likely that the maximum jacking force is the highest tendon
load that would be exerted on the deviators. The ratio of D/Do is equal to the deviation
saddle factor of safety. The strain gage plots typically indicated that at early load stages
the maximum strains occurred in reinforcement located directly above the tendon with the
highest deviation force, while at later load stages after a considerable amount of cracking,
the highest strains were in the link bar or loop bar reinforcement legs which were acting
in direct tension. The fracture locations in the legs of the reinforcement acting in direct
tension confirm this (see Fig. 10). For the modified specimens (4A, 4B, 5A, and 5B), it is
significant to note that all the strain gages indicated yield of reinforcement which revealed
efficent use of the reinforcement. Final failures were generally explosive and dramatic with
many of the deviator bars fracturing and splices opening (see Figs. 10 and 11).
For making strength comparisons, the critical load stages were denoted as microc-
racking, visible cracking, first yield of the reinforcement, and ultimate. Microcracking was
asumed to be indicated by the first apparent jump in strain indicated by the strain gages
(see Fig. 8). Visible cracking was noted when the first surface crack appeared. Yield of the
reinforcement was noted when any of the strain gages reached the yield strain, and ultimate
load stage was apparent due to the explosive nature of the failure. Magnitudes of these
critical load stages and the nominal design load D. are given in Table 3. All specim!'ns
which were intended to be properly detailed deviation saddles (all specimens except 2A
and 2B) had an acceptable factor of safety for ultimate load (values ranged from 2.24 to
3.16). The factor of safety was adequate for the yield load stage for the properly detailed
deviation saddles 1A, lB, 3A, 3B, 4B, and 5B (values range from 1.6 to 2.08). However,
for tests 4A and 5A, the factor of safety at the yield load stage was unacceptably low,
1.33 and 1.06 respectively. The factor of safety against visible cracking was adequate for
specimens 1A, 1B, 3B, 4B, and 5B (1.3 to 2.03). However, it was marginal for specimens
3A, 4A, and 5A (0.78 to 1.03).
262 Beaupre et al

Recommendations will be made which should remedy these deficiencies at the

visible cracking stage and at the yield load stage. Average factors of safety for the critical
load stages of all the tests except 2A and 2B are shown in Fig. 12.
The epoxy coated reinforcement had adverse effects on the behavior of the devia-
tion saddle at microcracking and visible cracking stages. The average reduction in strength
at the serviceability levels of microcracking and visible cracking was 17% and 24% respec-
tively. However, at the critical strength stage of yielding, coated reinforcement had little
effect on the behavior. At this stage, all the load is basically transferred to the reinforce-
ment which if well anchored is not particularly dependent on the local bond characteristics
of the reinforcement. The well anchored coated reinforcement favorably affected the behav-
ior of the deviation saddle at the ultimate load stage with an average increase in strength
of 18% since it allowed for the complete mobilization of all the reinforcement within the
deviation saddle. It can be concluded from these comparisons that the use of epoxy coated
reinforcement resulted in an increased redistribution of force within the deviation saddle
before fracture of the direct tension reinforcement. The comparison between uncoated and
coated reinforcement types is summarized in the bar graph of Fig. 13.

ANALYSIS METHODS

Two separate analysis methods were investigated for each test. The first method
utilized simplified analysis models (direct tension model, shear friction model, and beam
model), and the second method utilized strut-tie analysis models (tie model for direct
tension reinforcement, and strut-tie model for top surface reinforcement). These analyses
models were formulated based on the physical behavior of the specimens. The ¢> factor used
in comparisons with test results for both analyses had a value of one since the material
strengths and specimen dimensions were known accurately. The analysis of the direct
tension reinforcement was the same for both methods.
In the simplified analysis method illustrated in Fig. 14, the direct tension model 1 is
used for the analysis of the direct tension reinforcement in the deviation saddle. The shear
friction model 2 explains the actions of the shear friction reinforcement which transfers
the shear across a crack interface which may form below the tendon ducts. The beam
model3 explains the action of the top surface reinforcement which provides added strength
to the deviation saddle to resist pull-out forces. This reinforcement is stressed like tensile
reinforcement in a beam, and it also distributes surface cracks.
For the second method referred to, two separate strut- tie models, shown in Fig. 15,
were developed. One models the contribution of the primary direct tension reinforcement
and the other the contributions of the top surface reinforcement. The strut-tie model is
based on the premise that reinforced concrete structures carry load through a set of com-
pressive struts which are distributed and interconnected by tensile ties. The reinforcing
bars utilized in direct tension in the deviation saddles are simply tension ties linking the
deviation force to the box reinforcement. The strut-tie model for the top surface reinforce-
ment resistance is a combination of compressive struts branching from the average location
of the tendon duct to the tension tie which is the top surface reinforcement. Complete
details and results of the analyses are presented in Ref. 4. Comparisons of test results with
analytical predictions are included in Table 3. The values shown are at ultimate and the
analytical results are based on the measured ultimate strength of the reinforcement in or-
der to show the accuracy of the analysis method. Design would be based on reinforcement
yield points and introduce further conservatism.
 Prestressing 263

CONCLUSIONS

The following conclusions can be drawn based on the ten deviation saddle tests of
this investigation.
1) The safety of deviation saddles has been verified in this investigation. Properly
detailed deviation saddles will perform adequately under service load conditions
and have a sufficient factor of safety at ultimate.
2) Excluding the purposefully misdesigned specimens 2A and 28, all specimens except
specimens 18 and 38 exhibited adequate ductility and thus gave sufficient warning
of the impending explosive failure. The specimens which resisted the pull-out force
mainly by direct tension reinforcement (specimens 4A, 48, 5A, and 58) always
displayed adequate ductility because the reinforcement had to strain substantially
to fail.
3) The epoxy coated reinforcement had adverse effects on the behavior of the devia-
tion saddle at microcracking and visible cracking stages. However, at the critical
strength stage of yielding, coated reinforcement has little effect on the behavior.
The well anchored coated reinforcement favorably affected the behavior of the
deviation saddle at the ultimate load stage.
4) From the cracking patterns and the strain data, three behavioral mechanisms
were evident in the deviation saddle. They were the pull-out resistance of the
direct tension reinforcement, the flexural beam type resistance of the top surface
reinforcement, and the shear friction resistance of the specimens across the critical
cracked plane which was observed to be directly below the tendon ducts. The
pull-out resistance of the direct tension reinforcement and the effective bending
of the top surface reinforcement are additive. Some uncertainty exists concerning
the effectiveness of this top surface reinforcement at the yield load stage of the
deviation saddle. The top surface reinforcement is beneficial in controlling and
distributing cracks on the top surface of the deviation saddle.
5) All final failures (except with special specimen 28) were the result of the fracture
of the direct tension reinforcement. Test observations and analysis indicated that
in all of these tests shear friction did not appear to be critical to the failure of the
deviation saddle.
6) The capacity of the specimens could be determined by either the simplified anal-
ysis models or the strut-tie analysis models. Both analysis methods appear to
produce reasonable agreement with the test results, although both methods rely
on subjective assumptions for the analysis of the top surface reinforcement.
7) In comparing the two types of reinforcement which are utilized to resist pull-out
force in this study (direct tension reinforcement and top surface reinforcement), it
is obvious that the direct tension reinforcement (the link bar in specimens 1A, 18,
2A, 3A, 38 and the loop bar in specimens 4A, 48, 5A, 58) is significantly more
efficient than the top surface reinforcement in resisting the deviated force.
8) The critical force which acts on the deviation saddle is the deviation force con-
tributed by the tendon with the maximum deviation component. In this test
series, this is the force that was closely confined by the direct tension reinforce-
ment. In the early configuration (specimens 1A, 18, 3A, and 38), this was the
264 Beaupre et al

deviated force of the corner tendon which had both vertical deviation and horizon-
tal deviation. The other tendons of the early specimens did not greatly influence
thP deviator capacity of the specimens because they were not enclosed within the
critical reinforcement (direct tension reinforcement). In the revised configuration
(specimens 4A, 4B, 5A, and 5B), both tendons were enclosed in separate direct
tension reinforcement, but the corner tendon was more critical since it had greater
vertical deviation than the other tendon. The basic direct tension reinforcement
around the critical tendon should be proportioned for this maximum tendon devi-
ation force. The other tendons could be provided with the same reinforcement to
simplify detailing, or some lesser amount determined by a similar analysis based
on the individual tendon deviated force.

RECOMMENDATIONS

The following recommendations are made based on the ten deviation saddle tests
of this investigation. Recommendations are focused on the deviation saddle since it is the
weakest of the three basic types of tendon deviators. However, these recommendations can
be conservatively applied to the diaphragm and rib or stiffener since these type of deviators
generally have added strength contribution from the concrete.
1) Total service load design forces for the deviator should be the sum of the vertical
and horizontal components of the deviated force from each tendon. These can be
calculated as the maximum allowable initial jacking force multiplied by the size
of the angle change for the vertical and horizontal planes of the tendon. Under
AASHTO Specification5 , the maximum allowable initial jacking force is limited to
80% of the ultimate strength of the tendon (0.8(fpu)(Ap,)).
2) At service load levels, reinforcement stresses should be limited to the specified
allowable stresses in AASHTO Sec. 8.15.2.
3) For load factor design, neither AASHT0 5 nor ACI 1 1 clearly specify an appropriate
load factor for the prestress tendon force. In view of the explosive nature of
failure, and in order to guarantee a reasonable factor of safety commensurate with
other load and resistance factor combinations, it is suggested that for this specific
application the load factor on prestress force should be at least 1.7. Conventional
reinforcement should be assumed at the yield point. The t/J factor that should
be used in the design of the direct tension reinforcement should be 0.90 since the
primary acting force is tension. The t/J factor for the shear friction calculation
should be 0.85.
4) The recommended design detail is very similar to the modified deviation saddle
detail (specimens 4A, 4B, 5A, and 5B) of the test series shown in Fig. 4. The
general approach to the design of the deviation saddle should be to rely only
on the very efficiently utilized direct tension reinforcement for the pull-out force
resistance of the deviation saddle. Any contribution to the pull-out resistance from
the concrete is ignored as is any additional resistance from any beam type element
above the tendon ducts.
For the direct tension reinforcement, utilize small closed rectangular stirrups (la-
beled as loop bars) along the entire deviator tendon axis which loop around each
individual tendon duct of the deviation saddle and are well anchored in the box-
deviator corner node under the top mat of reinforcement of the bottom flange.
Additionally, provide closed tie reinforcement which encircles the entire corner
 Prestressing 265

node, ties the node into the web and flange, and provides reinforcement along the
deviator top surface for controlling and distributing surface cracks (this reinforce-
ment should be neglected in the calculation of the pull-out capacity). An amount
of closed stirrups of the same area and spacing as the loop bar reinforcement should
produce very satisfactory results.
Each individual loop bar group should be designed with the specified load factors
and tjJ factors to resist the full pull-out force of the tendon with the largest vertical
deviation. It will be more economical and will minimize fabrication errors to
provide the other tendons with the same reinforcement regardless of their tendon
deviation forces. The dimensions of the loop bars and the outer closed stirrup
must be based on the tendon duct curvatures and outer diameters, reinforcement
clearances, and development lengths. For the loop bars, the minimum clearance
at the highest point of the tendon duct should be approximately 1". The vertical
inside dimension of the outer closed stirrup should be at least 2" larger than the
loop bars. The maximum bar size utilized for the loop bars should be limited to
a deformed #5 size so as to be able to fully develop the 90° hook.
5) Deviation saddle geometry utilized should be a horizontal top surface with vertical
sides. This makes fabrication of the segments easier and provides the lowest height
deviation saddle possible, which is critical in shallow highway bridge structures for
clearance of deviated tendons from adjacent deviation saddles.
The concrete dimensions of the deviation saddle are controlled by the requirements
of the tendon duct curvatures and outer diameter, reinforcement clearances, and
cover requirements. The lowest point of the tendon duct above the top of the
bottom flange should be based on required clearance (1" to 2") for constructability
(protective sheathing placed on extension of tendon duct is generally used for
external post-tensioning). The transverse location of the ducts should be as close
as possible to the web wall since it is desirable to have as small of an eccentricity to
the web as possible to minimize bending moments in the bottom flange. The width
of the deviation saddle in the longitudinal direction of the bridge is dependent on
the spacing and amount of reinforcement (4" to 6" center to center reinforcement
spacing is recommended to allow constructability). Also, it is dependent on the
minimum radius that the tendon duct can be bent.
6) Provide a full bottom flange width deviation saddle as shown in Fig. 16 for curved
spans with small radii when all the tendons in a deviation saddle have large hori-
zontal deviations. For straight spans, when the horizontal tendon angles are small
(less than 3°) and the horizontal components are directed either into the web or
away from the web, it should only be necessary to use a deviation saddle which
is similar to those tested. It may also be a advisable to provide the full bottom
flange width deviation saddle no matter what the tendon deviations are when
epoxy coated reinforcement is being utilized because it was observed in this test
series that the specimens with epoxy coating reinforcement cracked at a much
lower load than that of the conventionally reinforced specimens (averaged 24%
lower). Since the reinforcement is being epoxy coated because of severe corrosion
conditions, it would be advantageous to go one step further and provide the full
bottom flange width deviation saddle which will substantially increase the factor
of safety against visible cracking.
266 Beaupre et al

A reduction in concrete volume for the full bottom flange width deviation saddle
could be made by reducing the longitudinal dimension of the deviation saddle by
one-half in the center of the bottom flange at a certain distance from the tendon
ducts. Near the webs the deviation saddle would be the same as the models tested,
but would be joined to the opposite deviation saddle by a concrete strut half the
dimension of the deviation saddle.
7) In cases where a full bottom flange width deviation saddle is not provided, a
shear friction calculation should be made. In most cases, extra shear friction
reinforcement will not have to be provided since there is usually an excess of direct
tension reinforcement at the critical shear plane. Also, the outer closed stirrups
contribute to the shear friction reinforcement. The shear friction equation that
should be used for this check is Equation 8-10 of AASHTO Sec. 8.15.4.3. Any net
tension across the shear plane is taken into account by subtracting it from the total
capacity of the reinforcement crossing the shear plane. The maximum allowable
shear strength provided by this equation is the lesser ofthe two values, 0.09~Acv or
360Acv where Acv is the area of concrete section resisting shear transfer. This area
is assumed to be the area below the tendon ducts from the centerline of tendon
closest to the web wall to the front face of the deviation saddle. The mu factor is
taken as 1.4 since the deviation saddle is monolithically cast.

DESIGN EXAMPLES

Two design examples are presented to help clarify the recommendations discussed
above. The first one is shown in Fig. 17a and b for a deviation saddle from a straight span.
Since the horizontal deviations are less than 3°, a deviation saddle similar to those tested
is assumed. The second design example is shown in Fig. 18a and b for a deviation saddle
from a curved span. A full bottom flange width deviation saddle is assumed because the
horizontal angles are quite significant.

ACKNOWLEDGEMENTS

This paper is based on research sponsored by the Texas State Department of

Highways and Public Transportation and the Federal Highway Administration. In addition,
the basic testing facility and substantial financial support for the program were provided by
the National Science Foundation under Grant ECE-8419430. All opinions and conclusions
expressed are those of the authors and do not necessarily represent the views of the sponsors.
The authors would like to particularly acknowledge the advice and assistance of
Mr. Alan Matejowsky and Mr. Dean VanLanduyt of the TSDHPT who provided valuable
suggestions and insight leading to the development of the final deviator layout.
 Prestressing 267

REFERENCES

1. Carter, L.L., "Deviator Behavior in Externally Post-Tensioned Bridges," unpub-

lished Master's thesis, The University of Texas at Austin, August 1987.
2. Beaupre, R.J., "Deviation Saddle Behavior and Design for Externally Post-
Tensioned Bridges," unpublished Master's thesis, The University of Texas at Austin,
August 1988.
3. Powell, L.C., Breen, J.E., and Kreger M.E., "State-of-the-Art Externally Post-
Tensioned Bridges with Deviators," Center for Transportation Research Report
No. 365-1, The University of Texas at Austin, June 1988.

4. Beaupre, R.J., Powell, L.C., Breen, J.E., and Kreger, M.E., "Deviation Saddle Be-
havior and Design for Externally Post-Tensioned Bridges," Center for Transporta-
tion Research Report No. 365-2, The University of Texas at Austin, June 1988.

5. American Association of State Highway and Transportation Officials, Standard

Specification for Highway Bridges, Thirteenth Edition, Washington, D.C., 1983.
6. Treece, R., "Bond Strength of Epoxy- Coated Reinforcing Bars," unpublished Mas-
ter's thesis, The University of Texas at Austin, May 1987.
7. Hofbeck, J.A., Ibrahim, 1.0., Mattock, A.H., "Shear Transfer in Reinforced Con-
crete," ACI Journal, Vol. 66, No. 2, February 1969, pp. 119- 128.
8. Mattock. A.H., Hawkins, N.M., "Shear Transfer in Reinforced Concrete - Recent
Research," PCI Journal, Vol. 17, No. 2, March-April 1972, pp. 55-75.
9. Mattock, A.H., Johal, L., Chow, H.C., "Shear Transfer in Reinforced Concrete with
Moment or Tension Acting Across the Shear Plane," PCI Journal, Vol. 20, No. 4,
July-August 1975, pp. 79-93.
10. MacGregor, J .G., Reinforced Concrete-Mechanics and Design, Prentice Hall, New
Jersey, 1988.
11. American Concrete Institute, Building Code Requirements for Reinforced Con-
crete (ACI 318-83), Detroit, 1983.
12. Schlaich, J ., Schaefer, K., Jennewein, M., "Towards a Consistent Design of Rein-
forced Concrete Structures." PCI Journal, Vol. 32, No. 3, May- June, 1987.
13. Marti, P., "Truss Models in Detailing," Concrete Institute, Vol. 7, No. 12, Decem-
ber 1985, pp. 66-73.
268 Beaupre et al

Table 1 Deviator reinforcement properties

Yield Ultimate
Size Strength Strength Yield Strain
(ksi) (ksi) (micro in.fin.)
#1.25 36 51 1240
#1.5 45 63 1550
#2 45 63 1550

Table 2 Average 28-day compressive strength

Compressive
Specimen Strength (psi)
Thst lA 5650
Test 1B 5650
Test 2A 5700
Test2B 5750
Test 3A 6000
Test 3B 5800
Test 4A 5700
Test4B 5700
Test5A 5400
Test5B 5400

Table 3 Test results

Deviation Saddle Force (kips) (Nut est /Nucalc)

Specimen Nominal Microcracking Visible First Ultimate Simplified Strut

Design Cracking Yield Analysis* & Tie
(Do) Model
1A 10.0 - 13.0 16.0 23.3 1.04 0.92
18 11.2 13.7 20.3 22.9 26.9 1.01 0.87
2A 11.2 9.7 9.7 11.7 16.2 1.13 1.13
28 11.2 10.7 10.7 11.5 14.2 1.10 0.99
3A 10.0 - 10.3 17.3 28.0 1.19 1.04
38 11.2 12.5 14.8 23.2 25.2 1.30 1.13
4A 9.7 9.5 9.5 12.9 21.8 1.01** 1.01**
48 9.5 15.7 19.2 19.2 22.8 1.14** 1.11**
SA 9.7 7.6 7.6 10.3 24.6 1.25** 1.25**
58 9.5 12.5 14.1 18.3 25.0 1.21** 1.18**
Ave. 1.14 1.06

• Simplified analysis assuming beam element unrestrained at ends.

Based on capacity of critical tendon 2 which assumes no redistribution.
 External Tendons
in Grouted Polyethylene Sheathing

~
Diaphragm Anchorage .J ~
[fJ

~
[fJ
.. .
[fJ

~
Cfq

t-.:1
Figure 1. External Post-Tensioning in Long Key Bridge (From Ref. 1)
$
270 Beaupre et al

a) Diaphragm

b) Rib

c) Saddle
Figure 2. Deviator Types
 Prestressing 271

(#1.5 bar)
(4 req'd)

I' 11' 'I b

~~ 1. 7" .1

Figure 3. Specimens lA, lB, 3A, and 3B

272 Beaupre et al

(10 roq'd)

·[0
l7.:'f
(5roq'd)
,. I

t::F~
• 2.25" ..

Figure 4. Specimens 4A, 4B, SA, and SB

 Prestressing 273

Test Setup, Elevation (From Ref. 1)

N
.-11111...-
0 0 0 0 0 0 0 0 0

Test Setup, Plan (From Ref. 1)

Testing Concept (From Ref. 1)

Figure 5. Testing Apparatus

27 4 Beaupre et al

TEST 4A PHASE 1 AND 2

TOTAL VECTOR FORCE ON DEVIATION SADDLE
25 --0

20
en
Cl..
::;::: 15
.........
w
~ 10
0
u..
5

0
..,
... ... ..,~ ~ 0
"'
ot) ot)
-..., w w
w
N
w
N
w
N

"'~ ~"' """' ~

Ill Ill
~ ~" "'~
Ill
~ "'
~ ~ ~
til ~
~
N

~
~
CL
2
~
~

TEST 4A PHASE 1 AND 2

TOTAL VECTOR FORCE DIRECTION
70
,....
(/)
w
~ 65
0
0
w /eo
'-' 60
...Ill. j
z ~

0
i= lJ""""
~55
0::::
Ci
50

w
ot)

~
-... ~
ot)

~
~ ~
0
N
w
ot)
N
w
"'
N

~ ~ ~
"'~ "-0

~
C)

~
~ ~
(/) V>
0
(/)

N
V>
~
~
~
~
~ 2
~
D.
D.

Figure 6. Test 4A Loading History

 TOP FACE

:' , . ! LS 12.
13 t.·(-~
.· .· - 7
LS . ,.,,.......•' .'''....
,,,.,,
2 '.' '.
$ .1 }
., .D ..
.
1
.
...,.·.;,.
···::;:..: :·:: .. ',.. i'{>> LS 23

NORTH FACE ("A" SIDE)

LS 7-FIRST CRACK

SOUTH TO "8" END

·•llllllllonllo InIii<, ., .,··,·'i:$·' . . j , :< 1 11111111111111•· NORTH TO "A· END

_...__ PHASEI
--.....__PHASE/I
~
(/J
......
,..,
~
(/J
TOP FACE (/J

s·
(fq

Figure 7. Test 4A Crack Patterns 1:-..:l

--.}
C/1
 t-..?
TEST 4A PHASE 1 -....]
0\
DEVIATION SADDLE FORCE VS. REINFORCEMENT STRAIN
30 -a- "A" GAGE
t:::O

25
I -H-

--9--
"8" GAGE
"C" GAGE
en
~
.:
'ij

~ Ey _..,_ "D" GAGE

~
en
......
~zo
I ~ "E" GAGE e:..
ln ~ "F" GAGE
a...
::L
'-"
w 15
I --&-- "G" GAGE
-er- "H" GAGE
u
a:: -+- "I" GAGE
0
u... 10
-,-~.
~i-~F"
"A" "0" "G"
\ " ·s·J·c·Q·E""F"0"H"
5
I closed stirrup "I" loop bars

0+-----~--~------~--~----.----.-----.----.-----,
0
0
0
0
8 80 8 80 8 8 0 80
~ ~
ll) ll) ll) I()
I N P"l '<!"

STRAIN (MICRO lN./IN.)

Figure 8. Test 4A Reinforcement Strain Data - Phase 1

 TEST 4A PHASE 2
DEVIATION SADDLE FORCE VS. REINFORCEMENT STRAIN
30 -a- "A" GAGE
I
2~
r-Ey -
-
"8" GAGE
"C" GAGE
I
I
...J....__ ----
LULT -+- "0" GAGE
--+-- "E" GAGE
,.-...zo
(/)
a.. --+- "F" GAGE
~

::::15
(.)
- - - "G" GAGE
-m- "H" GAGE
0:::: -+- "I" GAGE
0
lL. 10 _ _.OF

,~~~~"'
"A" "0" "G"

5
\ ·slc·Q·E··F·Q·w
closed stirrup "I" loop bars

~
rJl
0+----+----~---,--~~----r----r----~---T----, .......
0 g g g ~
8 8 0

~
~ ~
80 0
~
8 rJl

I
0

""
~

""
~
n
0
-.:t
s·
rJl

crq
STRAIN (MICRO lN./IN.)
Figure 9. Test 4A Reinforcement Strain Data - Phase 2 t...:l
-.)
-.)
278 Beaupre et al

"A" Side

EBFracture location

\
EB

I fi [J AIII"P' tailed

\
EB

I 0 0
\e I 0 0
\
EB

I 0 0
0 0
Tendon 2
0 Tendon 1

"B"Side (Refer to Figure 2.14b)

Figure 10. Test 4A - Reinforcement Failure Locations

 Prestressing 279

Figure 11. Test 4A - Before and After Failure

 t-v
00
0
FACTORS OF SAFETY AT CRITICAL STAGES
RATIO D/Da FOR PROPERLY DETAILED DEVIATIO~ SADDLES O::i
('!I

3.------------------------------------------------, §
'i:l
~
~ AVERAGE ('!I
.......

2 ·+---- e.
0
0
',,
0

0
C) C)
_,
0 w
z z w ~
.:::.c ~ ~
u u >-
~ ~ ~
::::::>
u u
0
a:::
u
_,
w
m
::;'E Vi
5
Figure 12. Average Factor of Safety for Critical Stages
 COMPARISON OF REINFORCEMENT TYPES
RATIO D/Do FOR PROPERLY DETAILED DEVIATION SADDLES

3 1.18*
~ AVG. UNCOATED 2.78

~ AVG. EPOXY 0. 97*"

2 0.76*
0 0.83*
0
~
0

()
0 ~ 0 w
z z _J
~
sz
u
<{
~
u
w
>= ~
i=
~
C/l
f1:_ _J .......
0:::
u u ~ ~
*Ratio of epoxy coated reinforcement specimen strength to C/l
0 C/l
w
u
0::: _J
m
the uncoated reinforcement specimen strength s·
crq
:::;?; in
>
1:-.:)
Figure 13. Comparison of Factors of Safety for Reinforcement Types
co
......
282 Beaupre et al

Closed Stirrup

Open Stirrup
a) Early configuration

b) Modified Configuration Stirrup

Figure 14. Simplified Analysis Models

 Prestressing 283

\
\J

~F

c~~ r "'

Link
fR
};ill
-~
/ '-../

bar ' •I L [5 \.

Tension Ties Linking Deviation Force to Box Reinforcement

Compression Struts

Tension Tie

Strut-Tie Model for Top-Surface-Reinforcement Resistance

Figure 15. Strut-Tie Models

 ~
co
,.j;::.

t::l:l
Cb
~
"t:l
~
Elevation Cb
.....
e..
Strut half the width of deviation saddle

Plan View

Fi,gure 16. Plan and Elevation of Full Bottom Flange Deviation Saddle
 Prestressing 285

Example Pesjgn Ca!culatjon- Straight Span ·Two tendons

Tendon 2-19..0.5•0 270 ksi strands Aps-2.91 sq. ln. (closest to web waD)
Tendon 1-12-0.5·0 270 ksl strands Aps-1.84 sq. ln.
fc-6000 psi Grade 60 reinforcement

Maximum Allowable Jacking Force-0.8(fps)(Aps)

Tendon 2 •(0.8)(270 ksi)(2.91 sq. in.)-628.6 k
Tendon 1 •(0.8)(270 ksi)(1.84 sq. ln.)-397.4 k

Tendon Horiz. Oev. Vert. Dev. Horiz. Force Vert. Force

2
1
Total
+2.93.
+1.84.
+8.22·
o·
+32.1 k
+12.8 k
+44.9 k
+89.9 k
k
+89.9 k
o L (center of box)

Load Factor-1.7
0-0.9 (Tension)
0-0.85 (Shear)

oesjgn tendon loops based on 89 9 k

Fu-0(As)(fy)

As-Fut(0fy)-(89.9 k)1. 7/((0.90)(60.0)•2.83 sq. in.

Assume 1#4 bars As-0.20 sq. in.

I of loops-2.83 sq. inJ(2(0.20 sq. in.))-7.1 use 7-#4 loops each tendon
use 7-14 closed outer stirrups
Shear frictjon check,
Vu-(0Asfy-Nu)~ AASHTO Sec. 8.15.4.3

Area equal to two legs of 7 loops for each tendon and one leg of 7
outer closed stirrups

As·7(5)(0.20 sq. in.)-7.00 sq. in.

Nu-(89.9 k)(1.7)·152.8 k
~1.4 (monolithically cast)
fy-60 ksi
0-0.85
Vu(req'd)-(44.9 k)(1.7)•76.3 k

Vu(provided)-[(0.85)(7.0 sq. ln.)(60 ksij-152.8 k)1.4·286.9 k > 76.3 k O.K.

Vu(max provided)-360Acv-360.(15.5 in.)(28 in.}/1000.156.2 k > 76.3 k O.K.

Figure 17a. Design Example for Straight Span

286 Beaupre et al

Section A-A Maximum Deviated Tendon

Link bar vertical dimension=3. 75" +3.56"12+(12")(tan8.22")+2" +2(5/8")+ 1"+2(1/2")= 12.5"

(out-to-out dimension) use 13"

Link bar horizontal dimension=3.56"+(12")(tan 2.93")+ 1"+2(1/2")=6.2"

(out-to-out dimension) use s·

Outer stirrup vertical dimension (out-to-out)=13"+3"= 16"

~~a~:~
Development-90" hook
lhb =1200(db )/(f'c)112 =7./
ACI12.5.3.2 Factor=0.7(1oops only) 25"
I dh =7./(0.7)=5.4" uses· O.K.

Deviation Saddle Elevation

Figure 17b. Design Example for Straight Span

 Prestressing 287

Example Oesjgn Calculation- Curyed Span -Two tendons

Provide fun bottom flange width deviation saddle since sman radius
curved span

Tendon 2-19-0.6•0 270 ksi strands Aps-4.09 sq. in. (closest to web wall)
Tendon 1-19-0.6.0 270 ksi strands Aps-4.09 sq. in.
rc-6000 psi Grade 60 reinforcement

Maximum Allowable Jacking Force-0.8(fps)(Aps)

Tendon 2 •(0.8)(270 ks~(4 . 09 sq. in.)-883.4 k
Tendon 1 •(0.8)(270 ks~(4.09 sq. in.)-883.4 k

Tendon Horiz. Oev. Vert. Oev. Horiz. Force Vert. Force

2
1
Total
+,-4.o·
+,-4.o·
.a.o·
+S.o·
+,-61 .8 k
+,-61.8k
+,-123.6k
+122.9k
+92.3 k
+215.2k
L+ (center
of box)

Load Factor-1 .7
0-0.9 (Tension)
0-0.85 (Shear)
Oesjgn tendon loops based on 122 9 k
Fu-0(As)(fy)

As·Ft.V(0fy)·(122.9 k) 1. 7/((0.90)(60.0)-3.87 sq. in.

Assume 15 bars As-0.31 sq. in.

I of loops-3.87 sq. in/(2(0.31 sq. in.))-6.2 use 7- IS !oops each tendon

use HIS closed outer stirrups

No shear friction check required since tun bottom flange

deviation saddle provided

Figure 18a. Design Example for Curved Span

288 Beaupre et al

Section A-A-Maximum Deviated Tendon

Link bar vertical dimension=4.5"+4.31"/2+(12")(tana.0')+2"+2(5/a")+ 1"+2(5/a")=13.a"

(out-to-out dimension) use 14"

Link bar horizontal dimension=4.31"+(12")(tan 4.0' )+1"+2(5/a")=7.4"

(out-to-out dimension) use a·

Outer stirrup vertical dimension (out-to-out)=14"+3"=17"

~~
38"

LCJ
115 closed outer stirrup

I (lap- B"ll
~
Development-90' hook
lhb =1200(db )/(f'c)112 =9.7" 29"
ACI12.5.3.2 Factor=0.7 (loops only
ldh =9.7"(0.7)=6.a" use a· O.K.

Deviation Saddle Elevation

Figure 18b. Design Example for Curved Span

 SP 120-13

Externally Prestressed Concrete Slab

Bridges: Model Test Results
by C. Menn and P. Gauvreau

Synopsis: A research project is currently underway at

the Swiss Federal Institute of Technology, Zurich, to
establish the feasibility of an alternative structural
system for short-span highway bridges. Concerns over
the long-term durability of structural systems
currently used in the 25 to 40 m span range provided
the primary motivation for the study. The proposed
system consists of a solid concrete slab which is
externally prestressed. The external tendons are
deviated at the third points of each span using
struts. A 1:3-scale model bridge has been constructed
and is currently being tested to verify the behaviour
of the bridge under permanent, service and ultimate
static loads, as well as dynamic and fatigue loads.
The favourable results obtained thus far have
confirmed the feasibility of the proposed structural
system, and will serve as a basis for extending the
concept to spans greater than 40 m in length.

Kevwords: bridges (structures); CCl!TpOSite construction (concrete

and steel); concrete slabs; durability; models; prestressed
concrete; scale (ratio); serviceability; tests; unborrled
prestressing

289
290 Menn and Gauvreau

Christian Menn has designed over eighty bridges in

Switzerland, including some of the country's longest-
spanning w0rks. He holds a doctorate from the Swiss
Federal Institute of Technology in Zurich, where he
has been a professor since 1971.

Paul Gauvreau is a research associate at the Swiss

Federal Institute of Technology in Zurich. He
received his Master's degree in civil engineering from
Princeton University in 1983. He worked as a bridge
design engineer in the U. S. and Canada from 1983 to
1987.

DURABILITY AND THE STRUCTURAL SYSTEM

In recent years, the long-term durability of bridges

has become a focus of concern for engineers throughout
the world. An alarming number of bridges has fallen
into severe disrepair as a result not only of
inadequate inspection and maintenance, but also of
shortcomings in design and detailing. In addition to
the obvious task of repairing deterioration in
existing structures, it is also imperative that
engineers develop better ways of ensuring durability
in the design of new structures.

Adequate protection against deterioration results

from both careful detailing of individual bridge
components as well as proper overall design of the
structural system. The structural system can
significantly improve a bridge's resistance to decay
where the protection provided by detailing alone
cannot be relied on. For example, fully sealed
expansion joints in a multiple-span bridge help to
prevent damage in the concrete and bearings located
underneath. Leaking is possible, however, even in the
best detailed expansion joint. A higher degree of
protection is provided by eliminating as many of the
joints as possible through the use of a continuous
superstructure. A multiple-span bridge which is fully
continuous between abutments can therefore be regarded
as more intrinsically durable than a similar bridge
composed of simply-supported spans.

In a similar manner, proper design and careful

detailing of the superstructure cross-section can also
significantly increase the bridge's resistance to
decay. In the particular context of concrete
 Prestressing 291

structures, the ideal cross-section with regard to

long-term durability is one in which

the surface area exposed to harmful

environmental conditions is minimized,

the proper placement, vibration and curing of

the concrete are in no way hindered,

adequate cover for the reinforcement is

guaranteed, and

every concrete surface is accessible for

inspection and maintenance.

By evaluating the overall design of the section

according to the above criteria, it is possible to
establish the level of intrinsic durability of a given
cross-section type.

Solid slabs have the highest intrinsic durability of

any section type. For a given width of roadway, the
total exposed surface area of the section is
minimized. The arrangement of reinforcing and
prestressing steel in a slab bridge permits problem-
free placement and vibration of concrete. Specified
cover can likewise be easily achieved at all
locations. Access to every concrete surface is
guaranteed due to the absence of interior surfaces.
Solid slabs are, however, not normally economical for
spans greater than 25 m in length.

On the other hand, the intrinsic durability of box

girders, T-girders and other similar types of sections
depends to a large extent on the dimensions of the
section, and, consequently, on span length. The most
critical factors in this regard are web thickness and,
in the case of hollow sections, available interior
clearance. Thin webs lead to a congested arrangement
of reinforcing and prestressing steel which hinders
the proper placement and vibration of the concrete.
Insufficient interior clearance further complicates
the concreting operations. Good workmanship, which is
crucial to high-quality concrete, is very difficult to
achieve when workers are unable to stand erect. Lost
forms, which have been used in hollow sections of very
limited clearance, permanently prevent access to
concrete surfaces. Webs of sufficient thickness and
adequate interior clearance will normally only be
found in spans greater than 40 m in length.

The 25 to 40 m span range is therefore problematic

with regard to long-term durability. These spans are
too long for the economical use of solid slabs, yet
292 Menn and Gauvreau

too short for the durable construction of box girders.

The cross-sections traditionally used in this range of
spans, i.e. box girders, T-girders (including precast
girders) and voided slabs, all suffer from low
intrinsic durability as a result of thin webs and, in
the case of boxes and voided slabs, insufficient
interior clearance.

AN ALTERNATIVE STRUCTURAL SYSTEM

The Swiss Federal Institute of Technology in Zurich

is currently investigating the use of innovative
composite structural systems for highway bridges. As
a result of the low intrinsic durability of
conventional concrete structural systems in the 25 to
40 m span range, a major phase of the study has
focused on developing an alternative composite
structural system which would guarantee both economy
and durability in this range of spans.

External prestressing, in which unbonded tendons are

located outside of the concrete, can improve the
inherent durability of box and T-girders by relieving
the congestion of reinforcing and prestressing steel
in the webs. The benefits of external prestressing
are rather dubious, however, in the particular case of
box girders, since removing the tendons from the webs
in no way helps to solve the problem of concreting and
formwork removal operations under conditions of
insufficient interior clearance. The external
prestressing of boxes and T-girders was therefore not
given further consideration.

A more promising solution resulted from the use of

external prestressing in conjunction with solid slabs.
External prestressing offers an efficient means of
extending the economical range of slab bridges to
spans of up to 40 m. The high intrinsic durability of
conventional slabs, moreover, is in no way diminished
by external prestressing.

Figure 1 shows the proposed implementation of the

concept. The arrangement of slab, cables and struts
is admittedly not new; bridges of similar appearance
have been built as far back as the early nineteenth
century.

Clearly visible in Figure 1 are the concrete slab,

external tendons, and short structural steel posts
located at the 1/3 and 2/3 points of each span. The
posts deviate the external tendons and transfer the
vertical component of their force to the slab. The
 Prestressing 293

depth to span ratio of the slab, the dominant visual

element, gives an overall impression of great
slenderness.

The system is intended for multiple-span continuous

bridges and viaducts. Its design is therefore adapted
to a span-by-span construction sequence. Of primary
importance in this regard is the prestressing concept,
which, apart from ensuring good behaviour at both
serviceability and ultimate limit states, reduces
falsework requirements to a minimum. The tendons have
been divided into two distinct groups: the external
tendons, visible in Figure 1, and the internal
tendons, located within the slab.

The internal tendons are continuous for the entire

length of the bridge and are arranged in two parallel
groups. According to the proposed construction
sequence, they would be stressed immediately after
hardening of the concrete, in order to enable prompt
removal of falsework and formwork. The required area
of steel is thus designed to resist dead load during
the construction stage. After the bridge has been
completed, these tendons contribute primarily to the
slab's live load resistance in the secondary spans
between the piers and the short steel posts.

The external tendons are coupled to short tendons

which are cast into the slab over the supports. The
external tendons below the slab are divided into three
segments, visible in Figure 1: (1) a segment sloping
down from the slab to the bottom of the steel struts
in the left third of the span, (2) a horizontal
segment between the struts in the middle third of the
span, and (3) a segment sloping up from the struts to
the slab in the right third of the span. Each segment
is anchored at the base of the posts, thus eliminating
the need for deviation saddles and allowing the entire
tendon to be stressed at the base of the strut. After
completion of the bridge, dead load is resisted
primarily by the external tendons. The area of steel
is calculated to ensure safety of the completed
structure at ultimate limit state. The prestressing
force is chosen to provide a nominal upward camber
under permanent load, which results in an effective
prestress of about 45 percent of the yield stress. An
appropriately chosen combination of duct, grout,
grease or wax will protect these tendons from
corrosion. They are at all times accessible for
inspection and maintenance. They can also be easily
replaced, which further enhances the inherent
durability of the system.
294 Menn and Gauvreau

This prestressing concept makes possible a rational

span-by-span construction sequence, a typical cycle of
which is as follows:

1. Cast central portion of the superstructure on

falsework, excluding the cantilevers.

2. Stress internal tendons.

3. Remove falsework and formwork.

4. Stress external tendons over the supports.

5. Erect vertical struts, install and stress

external tendons (first stage) .

6. Cast cantilever portions of the section on

travelling forms, and stress transverse tendons.

7. Grout external tendons over the supports and

internal tendons.

8. Install waterproofing and wearing surface.

9. Stress external tendons (second stage) and

adjust bridge profile.

SCALE MODEL TEST PROGRAM

The structural behaviour of the proposed system has

been thoroughly investigated. A major component of
this work has been the testing of a large-scale model
bridge, to confirm the results of the analytical
models used, and to reveal further insights into the
bridge's behaviour. Because the system is made up of
relatively flexible elements, dynamic testing was also
required. In addition, the model was used to study
the behaviour of the structure in fatigue, with
special attention paid to the external tendon anchors
at the base of the steel posts.

The model is based on a prototype bridge of two 30 m

spans, continuous over the intermediate support. The
scale is 1:3, which results in two 10 m spans (Figure
2). The overall deck width of the model is 4 m. The
typical cross-section, shown in Figure 3, consists of
a concrete slab which is thickened to accommodate the
internal tendons. The prestressing tendons are made
up of 7 rom diameter wires with standard buttonhead
anchors. The internal tendons, located entirely
inside the slab, consist of two 9-wire tendons in each
 Prestressing 295

web. The external tendons consist of 2 tendons of 14

wir:es each.
The model was constructed following the span by span
sequence proposed above. Similitude of stresses
between model and prototype was therefore ensured.
For convenience, however, the entire concrete slab was
cast simultaneously on falsework; travelling forms and
transverse post-tensioning were not used. The
sequence of the dead loads of the central portion of
the slab, the cantilevers and superimposed dead load
was taken into account by loading the model with
concrete blocks at the appropriate times and
locations. These blocks, visible in Figure 4,
compensate for the reduced self-weight of the model,
making possible a 1:1 correspondence between stresses
in model and prototype.

Five principal aspects of the bridge 1 s behaviour

were investigated:

1. Behaviour Under Permanent Load

The bridge 1 s behaviour under the effects of dead

load, internal prestressing, and external prestressing
was investigated during the first phase of the test
program, to determine the long-term effects of creep
and shrinkage. During a ninety day monitoring period
following the completion of the bridge, regular
measurements were taken of the vertical deflections in
the slab and the forces in the external tendons.

The results of this phase are shown in Figures 5 and

6. The stress in the external tendons at the end of
the ninety day period was equal to about 86 percent of
the jacking stress. This value is consistent with
those obtained from conventional post-tensioned
structures. The net vertical slab deflection after
ninety days was equal to about 1 mm, or 1/10 000 of
the span length.

It can therefore be concluded that the effects of

creep and shrinkage on the proposed structural system
are not severe, and that the system 1 s long-term
behaviour poses no extraordinary design problems.

2. Dynamic Behaviour

This phase of the test program was undertaken to

determine appropriate impact factors for live load.
296 Menn and Gauvreau

It was conducted in two separate stages, whereby

natural frequencies and damping were measured both
before and after the application of service live load.
In this way, it was possible to determine to what
extent cracking in the slab influenced the dynamic
behaviour of the system.

The natural frequency of the bridge was obtained

from measurements of its amplitude in forced vertical
vibration as a function of forcing frequency. A 100
kg oscillating mass was mounted under the slab at the
midpoint of one of the spans. The amplitude of the
mass was held constant at 0.01 m, while the forcing
frequency was varied. The natural frequency of the
cracked system was approximately 2. 58 Hz. The
response curve of the uncracked system was almost
identical to that of the cracked system; the uncracked
natural frequency was approximately 2. 72 Hz. The
effect of cracking on the dynamic behaviour of the
bridge can therefore be neglected.

The natural frequencies measured for the model

corresponded to a frequency of about 1.5 Hz for the
prototype. This corresponds to a dynamic
amplification factor of 1.48 for a single truck (1).
For the more severe case of full live load on the
bridge, however, the impact factor is approximately
equal to 20 percent, which is consistent with impact
factors for conventional structural systems in this
span range.

3 Behaviour at Serviceabilit¥ Limit State

This phase of the test program considered the

combined effects of dead load, internal and external
prestressing, and service live load. Its purpose was
to gain insight into the structural system's behaviour
at serviceability limit state. The live load was
simulated using hydraulic jacks at six locations per
span. The maximum jack force, Qmax' was equal to 29
kN.

This was most extensive phase of the entire

experimental program. Each test consisted of loading
a given combination of the jacks up to full design
load, to simulate a desired live loading pattern.
Forces in the external tendons and vertical
deflections of the slab were recorded for all tests.
The principal live load cases are as follows:

full live load on both spans,

 Prestressing 297

full live load on one span only,

partial live load applied to one side of the

bridge centreline, to produce torsion,

partial live load applied to one half of one

span only (critical for longitudinal bending in
the slab).

Although not strictly a live load case, an

additional test was conducted to simulate the removal
of one of the external tendons, to confirm that the
tendons are indeed easily replaceable.

Figures 7 and 8 show the increase in external tendon

force and slab deflections for live load symmetrically
applied to both spans. It is evident from these
figures that the behaviour of the system is linear up
to maximum service load levels. The maximum increase
in external tendon stress is equal to about 8 percent
of the tendon's yield stress. (The tendon stress lies
well within the upper and lower bounds given by the
analytical lower and upper limits, respectively, of
slab stiffness) . The maximum measured increase in
tendon stress, occurring when live load is applied to
one span only, is equal to only 12 percent of yield
stress. The external tendon forces due to live load
are therefore relatively small.

4. Fatigue Behaviour

Of primary importance in this phase of the test

program was the behaviour of the external tendon
anchors, located at the base of the structural steel
struts. The design fatigue live load, equal to about
33 percent of full service live load, was simulated
dynamically by means of the same oscillating mass used
for the dynamic tests.

The model underwent a full 2 million cycles of

fatigue live load with no signs of distress anywhere
in the structure. It can thus be concluded that
conventional buttonhead prestressing anchors will be
adequate for the proposed structural system. The
reason for this favourable fatigue behaviour is due to
the low stress range in the external tendons, equal to
about 4 percent of the yield stress under design
fatigue load.

5. Behaviour at Ultimate Limit State

298 Menn and Gauvreau

The only test phase not completed at the time of

writing is the investigation of the model's behaviour
at ultimate limit state. The magnitude of the live
load will be increased beyond service load until
failure of the structure occurs. The purpose of this
test is to determine the ultimate load of the system
and the mode of failure. From this test, the inherent
safety factor against collapse will be determined. Of
particular interest will be the amount of plastic
hinging required in the slab before yielding of the
external tendons can occur.

FUTURE PROSPECTS

The model test program's excellent results have not

only confirmed the bridge's feasibility, they have
also laid the groundwork for externally prestressed
concrete slab bridges of greater span lengths. The
basic concept is by no means restricted to the 25 to
40 m span range; the intrinsic durability, good
structural behaviour and rational construction
sequence of the proposed structural system could also
be made available to bridges spanning up to 60 m.

The main challenge in adapting the concept to longer

spans will be to ensure adequate stiffness. This will
be accomplished by replacing at least part of the
prestressing steel in the external tendons by
structural steel, whose axial stiffness is roughly
five times greater than that of prestressing steel for
a constant yield force. In the case of a non-
prestressed lower support system, it would be
necessary to control long-term deformations by means
of fabrication camber and possible jacking of the
structural steel during erection. Stiffness
requirements during the construction stage will
require some modification of the proposed construction
sequence. It will be necessary to erect at least a
portion of the lower support system before removal of
the falsework.

Large shear deformations of the unbraced central

panel were effectively prevented in the 25 to 40 m
span range by the slab, which was sufficiently stiff
over the relatively short interior spans. These
deformations will be much larger for interior spans
approaching 20 m. The use of X-bracing in the central
panel will eliminate these undesirable deformations,
while still retaining the slender slab, a definite
visual asset.
 Prestressing 299

Another modification to the original structural

system which is being considered is the addition of
compression members between the bottom of the struts
and the intermediate piers, substantially increasing
the lever arm for resisting negative moment.
Preliminary studies have shown that this measure also
contributes somewhat to increasing the overall
stiffness of the system.

These enhancements to the original concept will be

investigated in further detail using the scale model
bridge, modified accordingly. We believe that this
system holds great promise for a new generation of
efficient , economical, and inherently durable
structures.

REFERENCES

1. Menn, C., Stahlbetonbr"t:tcken, Springer-Verlag,

Vienna, 1986, p. 67.

Fig. !--Externally prestressed concrete slab bridge

 C;.j
10.00 10.00
g
~ ·r ·r :~ t t ~

M
3. 33 3... 3.. . 3... 3.3. 3.. .

p;Aj · ~ · ~~
s=
~
::l
::l
~
::l
0...
CJ
~
SECTION .:
<
~
~

~~~~-,r,,----=---=..--,rr
I' 'I
--==.=;;- t'J -~---=--,.,.---==
I' ,
---,'11_.=-....=-_--,~,1
"
II
'I
.,L-=---=--=
'"
nl
' 'I IL•
hi
II
II
_jUL-==---=- _jLJL_-=_Jj --:-;
l_l=---=-=- _juL__==--_jUL_-=---=-
~ LJI
nl
IIL•I
lnl I'
_jr

1
PLAN

Fig. 2--Scale model: longitudinal section and plan

 Prestressing 301

r-

0
q ,.,,
I<

,.,,.,
,,-i$& z
~

"'en
p.

,.._
i_ II
II
-
l
l
<(
a..
~
•rl
~
0

;
<D (/) -rl
,.,C7l z
'-'
u
QJ
0 en
0 f--- ~- --- - ~ I-
-i z ~

,.._
,.,
C7l
0 "'
u
·rl
<D
• •••••••
II i= p.
:>,
II u '-'
L&J
(/) ~
..
•••••• ,....- QJ

,.,,., ., ~ .....
'"Cl
s
0

QJ
~

"'u
U)
I
r: I
0 <')
q
- , ., ()()
·ri

""'
[>
L

-
302 Menn and Gauvreau

Fig. 4--Scale model in laboratory

 Prestressing 303

100 ·•

--
0
Q.
98

96 \
\
-en
~
0 94

92
\._
.,.,
en
w 90
............1'--.
a:
1-
en
88
~.- ..._._1--·- ........ 1-o
86 •
84
0 10 20 30 40 50 60 70 80 90
TIME (Days)
Fig. 5--External tendon stresses under permanent load

-
2.5

E
- ;·
E 2.0 •~·...... ·-· ~ ........
.,
,................
z 1.5 • ~ .......
0
........\
1-
(.) 1.0 1/
I!
w
...J
~
u. 0.5
w
c
0.0 .•
0 10 20 30 40 50 60 70 80 90
TIME (Days)
Fig. 6--Slab deflections at midspan under permanent load
304 Menn and Gauvreau

12

-
--
10
>-
c.. ~
~ 8 ~~~
0
~'\)~
().'\) (S
en 6 -:.
en ~c
w 4
a: 0 '((12
1-
en ooo ~~'((\
2 - '\':>
~c-

~2 0.4 0.6 0.8

LIVE LOAD (xOmax)
Fig. ?--External tendon stresses under live load-both spans loaded

DISTANCE FROM END OF BRIDGE (m)

4 6 8 10 12 14 16 18 20

-E
-zE -5

-7.5 -·- 0.250

0 -o- 0.500
-10
1-
(.) -·- 0. 750
-12.5
w -o-1.000
...J
LL -15
w
c -17.5

-20

Fig. 8--Slab deflecti0ns under live load-both spans loaded

 SP 120-14

Behavior of 1/5 Scale

Segmental Concrete Girders
with External and Internal Tendons
by B. G. Rabbat and K. Sowlat

Synopsis. lhe behavior of two segmental concrete girders

incorporating external tendons was compared to that of a
similar girder with internal tendons. lhe girders were
31-tt-long and consisted of 11 match-cast segments. Test
variable was the location of the tendon ducts. In the first
girder, the ducts were embeddeo in the girder cross section.
The ducts of the secono girder were extern a 1 to ttle concrete
cross section except at pier segments and intermediate
deviation diaphragms. lhe thiro girder was similar to the
second except that portions of the external ducts were embedded
in a second-stage concrete cast. lhe segments incl uoed
multiple shear keys ana were dry jointed. All ducts were
grouted. l:.ach girder was simply supported over a 30-tt span
and loaded statically to destruction under a two point loao.
The first and third giroers attaineo their respective flexural
strengths predicted by the classic bending theory for
mono 1ithi c girders with bonded tendons. The second gi roer
exceeded the flexural strength predicted by the provisions of
the AASHTO Specifications for members with unbondeo tenoons.

Keywords: bridges (structures); girders; post-tensioninJ;

precast concrete; prestressed concrete; prestressing steels;
research; scale (ratio) ; segmental construction; structural
design; tests; unhanded prestressing

305
306 Rabbat and Sowlat

/ll.l l=e1low Basile G. Rabbat is Manager of Structural Codes

Engineered Structures and Codes, Portland Cement Association:
Skokie, Illinois. Dr. Rabbat fs a member of ACI Committee 215
Fatigue of Concrete and Secretary of ACI Colllllittee 318:
Standard Building Code.
ACI member Koz Sowlat is Manager, Dallas (Texas) office,
Construction Technology Laboratories, Inc. He was formerly
Associate, The Datum/Moore Partnership Consulting Engineers,
Dallas, Texas.
INTRODUCTION
This paper was prepared for the International Symposium on
External Prestressing to sulllllarize an experimental investi-
gation conducted at the Construction Technology Laboratories,
Inc. on girders with external tendons. More details about this
program are available in Reference 1.
During the last decade, use of external tendons has emerged
as an economic technique to post-tension segmental concrete
bridge girders in the United States and Europe. External
prestressing has also been used to rehabilitate and strengthen
bridge girders. Advantages of external prestressing have been
discussed by Virlogeux.2
OBJECTIVE
The objective of this investigation was to compare the
behavior of dry-jointed precast segmental girders with external
tendons with that of a similar girder with internal bonded
tendons. The effects of potential damage to anchorages at
girder ends due to a severe earthquake were to be eva 1ua ted
through releasing the anchor wedges of selected tendons prior
to loading the girders to destruction.
TEST PROGRAM
Test Specimens
The following three segmental girders were tested:
1. Bonded Tendon Girder with internal tendons
2. Unbonded Tendon Girder with external tendons
3. Modified Unbonded Tendon Girder similar to the second
girder except that a second stage concrete cast was
placed on the top of the bottom flange covering
portions of the external tendon ducts for bond
development.
Girder geometry and midspan cross sections are shown in
Fig. 1. Profile ot tendons is given in Fig. 2. At test time,
concrete compressive strength of segments ranged between 5810
and 7660 psi (40 and 53 MPa). All tendons were grade 270 low
 Prestressing 307

relaxation seven wire strands. More details about the test

specimens are given in Reference 1. The test setup is shown in
Fig. 3.
Test Procedure
Each girder was subjected to two loading cycles. In the
first cycle, each girder was loaded in small increments until
an inelastic behavior was observed and a midspan deflection of
approximately 3 in. (76 mm) was reached. The girder was then
completely unloaded. Anchor wedges for the top two strands
were burned at both ends to simulate anchorage loss due to a
severe earthquake. Burning of anchorages was a last minute
addition to the test program. During the second loading cycle,
each girder was loaded in increments up to destruction.
BEHAVIOR OF SPECIMENS
A plot of the midspan applied moment versus deflection for
the first and second loading cycles for each of the three
girders is given in Fig. 4. Peak moments (including effects of
girder dead load and loading hardware) and corresponding
midspan deflections for first and second loading cycles are
summarized in Table 1. After unloading the girders following
the first loading cycle, the anchor wedges of the top two
strands were burned. Strand strains were monitored during this
process. For the two girders with external tendons, strand
strains indicated a loss of stress in the top two strands
immediately after burning the anchorages. Loss of strand
stress was attributed to slip of strands at the anchorages.
With a loss of bond for the top two strands at the girder end
regions, the effectiveness of these strands became questionable.
This explained the loss of strength during the second loading
cycle for the Unbonded Tendon and Modified Unbonded Tendon
Girders as depicted in Fig. 4.
Bonded Tendon Girder
After burning the anchor wedges following the first loading
cycle, the top two strands remained bonded. Tendon ducts were
inside the section and were grouted. Transfer of prestress of
the top two strands occurred in the end regions of the girder.
The bonded tendon girder failed in a flexural mode. Distress
started with a localized crushing of concrete along the top
surface of the concrete at midspan. It was followed by strands
breaking which led to a loss of load carrying capacity.
Unbonded Tendon Girder
During the first loading cycle, after exceeding the
decompression moment, the dry joints in the constant moment
region started to open. Diagonal cracks occurred at the first
joints of the shear spans. These are the joints next to the
loading points. Upon unloading the girder following the first
308 Rabbat and Sowlat

loading cycle, the joints closed completely. As noted earlier,

after burning the anchorages of the top two strands, loss of
prestress was observed f.or these strands. The short length of
strand bonded within the pier segment (end segment) was
inadequate for transfer of prestress. Loss of prestress
resulted in decrease of the effective axial compression in the
member. Further, the shear strength of the girder was reduced
due to loss of the vertical component of prestress in the shear
span. The Unbonded Tendon Girder failed in a shear compression
mode.
Modified Unbonded Tendon Girder
Because of the secondary concrete cast placed on the bottom
flange of this girder there was a delay in opening of the dry
joints in the constant moment region. Diagonal shear cracks
occurred at the first joints of the shear spans as observed
with the Unbonded Tendon Girder. Following the first loading
cycle and burning of the anchorages, the top two strands
s1i pped at the girder ends. The fa 11 ure mode during the second
loading cycle was similar to that of the Unbonded Tendon
Girder. A shear compression failure occurred.
FLEXURAL STRENGTH ANALYSIS
Flexural strength of prestressed sections can b~ calculated
from eg\lations given in the AASHTO SpecificationsC3J or the ACI
Code. 14J Alternatively, principles of classic bending theory
can be applied to compute the flexural strength of girders with
bonded tendons. This theory states that plane sections remain
plane after bending. A computer-aided analysis was performed
to determine the flexural strength based on strain compatibility
and equilibrium of internal forces. For the Bonded Tendon
Girder, the calculated flexural strength was 7260 kip-in. (820
kN m). This compares with a flexural strength of 7320 kip-in
(827 kN m) COIIJP.14ted according to the AASHTO Specifications13}
and ACI Codei4J provisions for bonded tendons. For the
Unbonded Tendon Girder, if all tendons were assumed bonded for
analysis purposes. the calculated flexural strength according
to the classic bending theory was 7210 kip-in. (815 kN m). The
difference in calculated strength is because of the difference
in strand locations in the Bonded lendon versus Unbonded lendon
Girders. During the second loading cycle. the Bonded Tendon
Girder exceeded the calculated strength by 5 percent. The
Modified Unbonded Tendon Girder reached the calculated strength
for bonded tendons during the first loading cycle.
The AASHTO Specifications(3) provide the following equation
to compute the average stress at ultimate in unbonded tendons.
fsu* = fse + 15,000 psi
Cfsu* = fse + 103.4 MPa)
 Prestressing 309

where
fsu* = average stress in prestressing steel at ultimate
load, psi and
fse =effective steel prestress after losses, psi.
Based on this equation, the flexural strength for the
Unbonded Tendon Girder was calculated at 4380 kip-in. (495
kN m) assuming six effective strands. Therefore the Unbonded
Tendon Girder exceeded the calculated strength during the first
loading cycle by 29 percent.
CONCLUSIONS

Based on the test results the following conclusions were

drawn.
1. The Bonded Tendon Girder and the Modified Unbonded Tendon
Girder reached the flexural strength predicted by the
classic bending theory for bonded tendons.
2. The strength attained by the Unbonded Tendon Girder
exceeded the fl exura 1 strength ca 1cul a ted according to the
provisions of the AASHTO Specifications for unbonded
tendons.
ACKNOWLEDGMENTS
The experimental investigation reported in this paper was
sponsored by the South Carol ina Department of Highways and
Public Transportation. The test specimens were designed by
Figg and Muller Engineers, Inc. of Tallahassee, Florida. The
test specimens were manufactured and tested at the Construction
Technology Laboratories, Inc., Skokie, Illinois.

REFERENCES
1. Sowlat. Koz and Rabbat, Basile G., "Testing of Segmental
Concrete Girders With External Tendons," Journal of the
Prestressed Concrete Institute, V. 32, No. 2, March-April
1987, pp. 86-107.
2. Virlogeux, M., "External Prestressing," IABSE Proceedings,
International Association for Bridge and Structural
Engineering, P-62/82, Zurich, Switzerland, 1982, pp.
101-108.
3. Standard Specifications for Highway Bridges, American
Association of State Highway and Transportation Officials,
Thirteenth Edition, Washington, D.C., 1983.

4. ACI Committee 318, "Building Code Requirements for

Reinforced Concrete (ACI 318-83)," American Concrete
Institute, Detroit, Michigan, 1983, 111 pp.
310 Rabbat and Sowlat

Table 1 - Data SWIIIIary

First Second
Loading Cycle Loading Cycle
Maximum Residual Maximum
Maximum Deflec- Deflec- Maximum Deflec-
Test Girder Moment* tion tion+ Moment* tion
(kip-in.) (in. ) (in. ) (kip-in.) (in. )

Bonded Tendon 6,910 2.95 o. 13 7,600 7.68

Girder
Unbonded Tendon 5,670 2.65 0.05 5,360 4.31
Girder
Modified Unbonded 7,250 3.26 0.18 5,800 4.34
Tendon Girder

*Includes moment due to girder dead load and loading hardware

+Upon unloading
Metric Equivalents: 1000 kip-in. = 113 kN m; 1 in. = 25.4 mm
 Prestressing 311

A-,s-,A-,
II
2 4
·- I Ill I II II II II
. ..
+0041 s041 A.-I
12'-o'j Nine Segments at 3'- o"
31'- o"
I
.·-
12'-0"
:

4'-o" 4'-o"

I
11fl
2'-4''

I
I
I
14 1'-4"

Section A-A
.I I.
Section B·B
Bonded Tendon Girder

LeQend
e 0.6-1n.01a Strand
• 0.5-in 01o Strand

Section B-B Section B·B

Unbonded Tendon Girder Modified Unbonded Tendon Girder

Fig. 1 - Girder Gearetry and Tendon I.a::ations at Midspan

312 Rabbat and Sowlat

15'-6"

(a) Bonded Tendon Girder

I in.= 25.4 mm

(b) Unbonded and Modified Unhonded Tendon Girders

Fig. 2 - Tendon Profiles

 Prestressing 313

Fig. 3 - Test Setup

314 Rabbat and Sowlat

8
Cycle 2 Bonded Tendon

6 rMod1f1ed Unbonded Tendon

Cycle I

J.'\.-,r-Unbonded Tendon
Applied 5
Moment, -v.A.!:-")J
1000 kip -in.
4
II
3 )1
2 /-/' Metric Equivalents:
// 1000 kip-in.= 113 kN.m
/ / lin.= 25.4mm

/~·
/•

Deflection , in.

Fig. 4 - Applied M::mmt versus l\tidspa.n Deflection

 SP 120-15

Strength and Ductility of a Three-Span

Externally Post-Tensioned
Segmental Box Girder Bridge Model
by R.J.G. MacGregor, M.E. Kreger, and J.E. Breen

Synopsis. An experimental investigation was conducted to examine the service and

ultimate-load behavior of segmentally precast box-girder bridges with external post-
tensioning "tendons. A primary interest of this study was to examine the effect of joint
type (dry versus epoxied joints) on the stiffness, strength, and ductility of the structure.
A three-span reduced-scale segmental box girder model was constructed, then tested in
three stages. Flexural behavior was examined first, then shear tests were conducted on the
damaged structure. Test results and observations are presented.

Keywords: box beams; bridges (structures); ductility; models;

post-tensioning; precast concrete; prestressing steels; segmental
construction; strength; structural design; tests; unborrled
prestressing

315
316 MacGregor, Kreger, and Breen

ACI member Robert J.G. MacGregor is a Senior Engineer for Arvid Grant and
Associates in Olympia, Washington. He received BS and MS degrees from the University
of Alberta in Edmonton, Alberta, Canada and a Ph.D. degree in Civil Engineering from
The University of Texas at Austin in 1989.
ACI member Michael E. Kreger is an Assistant Professor of Civil Engineering at The
University of Texas at Austin. He received BS, MS and Ph.D. (1983) degrees from the
University of Illinois at Urbana-Champaign. He is a member of ACI Committee 318-C,
Analysis, Serviceability and Safety, ACI-ASCE Committee 352, Joints and Connections,
and ACI Committee 215, Fatigue of Concrete. He has conducted research in the areas of
repair and strengthening of concrete structures, behavior of reinforced concrete buildings
subjected to seismic loads, and behavior of prestressed concrete bridge structures.
John E. Breen, FACI, holds the Nasser I. Al-Rashid Chair in Civil Engineering at The
University of Texas at Austin. He is the past Chairman of ACI Committee 318, Stan-
dard Building Code, and is a former chairman of ACI-ASCE Committee 441, Reinforced
Concrete Columns, and of ACI's Technical Activities Committee

INTRODUCTION

A significant number of precast segmental box girder bridges with external pre-
stressing tendons have been constructed in the United States and Europe. Substantial
economic savings have been indicated for this type of construction. However, questions
have been raised as to how these bridges will beha,ve when subjected to loads greater than
service-level loads. It is not known to what degree the behavior of these bridges will re-
semble the behavior of monolithic, fully bonded or unbonded prestressed concrete girders.
In addition, segmental box- girder bridges have been constructed with epoxy between the
match-cast segments (epoxied joints) and also without any type of joinery material in the
joints (dry joints). It is not known whether epoxied or dry joints have any influence on the
stiffness, strength, or ductility of the structure.
A one-quarter scale three-span externally post-tensioned box girder bridge model
was constructed in the Ferguson Structural Engineering Laboratory at The University of
Texas at Austin. The model was load- tested to determine the level of strength and
ductility that may be expected for precast segmental bridges with external tendons, current
tendon anchorage and deviation details, and alternate joinery details. A description of the
experimental program, evaluation of selected test results, and conclusions follow.

TEST PROGRAM

Description of Model Structure

The model bridge, shown in Fig. 1, was a three-span segmental box girder with
geometric symmetry about the center of the bridge. Figure 2 shows a plan view and
elevation of the structure. Each span contained ten typical segments, and over each support
was a pier segment containing the anchorages for the prestressing tendons. Because the
typical segments were match cast separately from the pier segments, a cast-in-place closure
 Prestressing 317

strip was provided between the typical segments and the pier segments at the ends of each
span. Joints between typical segments were either epoxy filled or dry (no epoxy). Both
the south and center spans contained epoxied joints.
Cross sections for typical segments and pier segments are presented in Fig. 3.
The typical segment shape was chosen to give a span-to-depth ratio (18.75) and efficiency
rating (0.60) typical of contemporary construction. A modified box section was developed
for use in the model structure to provide access to the external tendons. Webs were shifted
towards the center of the box to facilitate moving the draped external tendons outside the
box. At midlength of each typical segment was a full-height diaphragm through which
external tendons passed freely or were deviated as required.
Shear keys were used on segment faces to transfer shear across segment joints
and to provide an interlock between match-cast segments. Keys were also provided in the
flange regions, with their primary purpose being to assist in aligning the segments during
erection. Details for the shear keys are shown in Fig. 4.
Concrete strength for typical segments was nominally 6000 psi at 28 days. In
order to adequately resist anchorage forces in the pier segments, a concrete mix design
with a minimum 28-day compressive strength of 10,000 psi was used.
A schematic of the post-tensioning tendons is shown in Fig. 5. Tendons were
draped downward from high over the supports to just above the bottom flange near
midspan. Theoretical tendon locatio11.s are shown for sections at the exterior support,
midspan, and interior support. The structure was erected on traveling falsework using
the span-by-span method of construction. Single-span tendons (1A, 1B, 2, 4A, and 4B)
which contained 5-3/8" diameter strands on each side of the box, were used during erection
of individual spans. Additional multi-span tendons (3 and 5) containing 2-3/8" diameter
strands on each side of the box, were added after erecting span 2 and span 3. External
tendons were pressure grouted in their plastic and metal pipe sheaths at deviation points.
In addition to the external tendons, internal tendons were provided in the top
corners of the box and at the ends of the thin top flange. The additional tendons were
provided to augment flexural and torsional capacity, and to control shear lag. All internal
tendons had a straight profile and were anchored at the extreme ends of the structure.
Each internal tendon consisted of two 3/8" diameter strands. All prestressing strand used
in the model structure was Grade 270 low-relaxation strand.

Instrwnentation

The reduced-scale bride model was instrumented to measure reactions, deflec-

tions, and local deformations such as tendon strains and joint openings.
Reactions were measured at three of the four supporting piers. At each location,
two load cells were used to measure the reaction beneath each of the webs. Measured
reactions were used to provide a check of static equilibrium, and to provide information
about load distribution in the continuous structure.
Vertical deflections were measured at each support and at three locations in each
span using displacement transducers and dial gages. Horizontal displacements were also
monitored at supports and at midspan of the span being tested.
External tendon strains were measured using resistance-type strain gages at-
tached to individual strand wires at nine locations on each side of the three-span structure.
318 MacGregor, Kreger, and Breen

Strains were measured in each tendon in the inclined portions near the supports and also
in the horizontal region at midspan.
Distortions along the height of critical joints were measured using displacement
transducers and mechanical crack-width gages. Displacement transducers were attached
to the tension flange of the box, and spanned across the joint. The crack monitors placed
across bottom-opening joints were located at the top and bottom of the vertical portion of
one of the webs and on one end of the bottom flange. For top-opening joints, the inclined
tendons did not allow access to the web regions, so a single crack monitor was attached to
the top flange.

Load Tests

A series of weights to compensate for model effects were applied during construc-
tion to bring the model to the correct dead load configuration.
The three-span model was then loaded to examine the complete range of flexural
behavior for the dry and epoxy-jointed exterior spans. The model was also subjected to
very low-level torsional loads and to loadings that were intended to mobilize the shear
strength of the dry and epoxy-jointed spans. The test program consisted of three distinct
load phases:
• structural characterization
• factored loads
• ultimate strength
The first phase of testing involved loading the structure to the design service live
loads and then increasing loads to higher levels to establish the cracking loads at critical
locations, or the decompression loads at critical joints or existing flexural cracks throughout
the structure. In the second phase of testing, factored design loads were applied to the
structure. In the final phase, the structure was loaded until the ultimate strength was
attained. Failure of both the north and south span (dry vs. epoxy joints) was dominated
by flexural behavior. Exploratory tests were then carried out on the damaged structure to
investigate shear behavior at opening joints.
Each possible load configuration was analyzed using a limit analysis to determine
the location of critical joints. The flexural and shear- test load locations were chosen so
that the same joints were critical for both tests. If two different joints were critical for
the flexure and shear tests, the desired critical shear mechanism may not have developed
properly in the cracked epoxy-jointed span.
So that comparisons could be made with the AASHTO HS-20 truck load, it was
necessary to determine the quantity of applied load that was equivalent to the reduced-
scale service live loads. Because the tests were dominated by flexural behavior, moments
at joint locations were used to convert between service loads and test loads. The applied
"equivalent live load with impact" (LL+I) was chosen to provide the same moment at the
critical joint as the maximum service-load moment calculated at any location. Maximum
service-load moments were determined using influence diagrams developed for each joint.
A schematic illustrating the positions of loads during flexure and shear tests on
exterior spans is shown in Fig. 6.
 Prestressing 319

TEST RESULTS

Service Load Behavior

The measured deflected shape of the three-span structure is shown in Fig. 7

for a typical service live load application on the dry-jointed north span. The measured
maximum service live load deflections, as determined in different load cycles, were 1/5660
for the dry-jointed exterior span, 1/6250 for the epoxy-jointed exterior span, and 1/7500
for the epoxy-jointed interior span. The deflection in the dry-jointed exterior span was
approximately 10 percent more than for the epoxy- jointed exterior span. This difference
may be caused by a slightly smaller effective cross-section in the dry joints caused by
differential shrinkage in the thin flanges of the precast segments.
The live-load tendon stress increases in the midspan region of the load span were
measured to be less than 2 ksi in all spans. The stress response remained constant for
five consecutive live load cycles indicating that the tendons did not slip at the deviators at
service level loads.
Although the live-load stress range was small and slip was not apparent during
live load cycles, there is need for research to assess the fatigue properties of external tendons
at the deviation locations. The change in tendon force between two adjacent segments of
an external tendon occurs 'by friction while under-going a concentrated angle change at
the deviators (Fig. 8). The force transfer occurs over a short length under high lateral
deviation pressures. The friction force combines with the lateral pressure to induce a high
surface shear on the strand wires that are in contact with the deviation hardware. Also,
repeated load cycles or the occurrence of a previous overload may increase the potential
for slip at the deviators. The high lateral pressures and surface shears at the deviators and
the potential for tendon slip indicates a possible fretting fatigue problem (1 ).

Cracking Load vs Decompression Load

A primary purpose for using epoxy at segment joints is to provide reserve capacity
against joint opening for over-load conditions. In the epoxy jointed exterior span of the
model, cracking occurred through the concrete adjacent to a midspan match-cast joint at
an applied load of 5.4 multiples of live load including impact (5.4*(11+I)). For subsequent
applied load cycles, the load required to decompress the flexural tension fiber, and cause
the crack adjacent to the epoxy joint to begin to open, was measured at an applied load of
2.6*(11+I). This measured decompression load for the cracked south span was somewhat
higher than the measured decompression load of 1.9*(11+I) for the dry jointed north
span. However, this was affected more by difference in effective prestress than joint type.
If zero tension is used as the limit for service behavior, then the epoxy joints provided
a potential factor of safety against joint opening of approximately 2.0. Similar behavior
was also noticed in the epoxy jointed interior span. In setting design criteria, however, it
should be realized that the true factor of safety might be less than this because of traffic
overloads, calculation inaccuracies, actual insitu epoxy behavior, and fatigue behavior of the
concrete/epoxy joint. It would therefore be prudent to specify a small residual compressive
stress in the extreme flexural tension fiber for epoxy-jointed segments without bonded
reinforcerr.ent crossing the joint. In dry joints, without bonded reinforcement crossing the
joint, the beneficial tensile capacity offered by the epoxy is not present, so higher design
residual compressive stresses are recommended.
320 MacGregor, Kreger, and Breen

Factored Loads

After completing the service load tests the three-span structure was loaded with
additional weight to simulate the factored dead load condition of 1.3*DL. Each of the
exterior spans of the structure was then individually loaded with the factored design live
load with impact, 2.86*(LL+I). Factored load tests were not conducted on the interior
span. The structure behaved linearly throughout the load cycle with a slight reduction in
stiffness when the midspan joint decompressed. At these higher load levels the measured
maximum factored live load deflections, as determined in different load cycles, were L/1764
for the dry jointed exterior span and L/2310 for the epoxy jointed exterior span. In this
case the deflections in the dry jointed span are approximately 25 percent more than in
the epoxy-jointed exterior span, with the difference caused by the reduced effective cross-
section in the dry joints and the tensile capacity in the uncracked regions of the epoxied
joints.
The factored-load tendon stress increases in the midspan region of the loaded
span were measured to be less than 5 ksi in all spans. The tendons did not appear to slip
at the deviators for any of the factored load cycles.

Mechanism Behavior

The applied load is plotted against the resultant midspan deflection for the ulti-
mate load test of the dry jointed north span, in Fig. 9. The deflections represent the net
deflection of the structure, at the location shown on the schematic, after adjustment for
support deflections. The deflections increase linearly with applied load up to the decom-
pression load, Pd. As the midspan joints begin to open, stiffness reduces, and deflections
increase at an escalating rate. The stiffness continues to decrease until the support joint
opens and a mechanism forms. For load levels higher than the "mechanism load", Pm,
the stiffness remains relatively constant with slight decreases as the ultimate strength is
approached.
The measured deflected shapes of the three-span structure with factored dead
load (1.3*DL) and increasing levels of applied load are shown in Fig. 10. At the applied
service live load, l.O*(LL+I), and the applied factored design load, 2.9*(LL+I), the deflec-
tions are small and the deflected shape appears as a smooth curve. The deflected shape
remains smooth until the midspan joints open widely at 3.0*(LL+I). Beyond this load,
"hinging" occurs at the midspan joints, and the midspan deflections increase considerably.
When the support joint opens at 4.8*(LL+I) the mechanism forms and deflections begin to
increase very rapidly. Due to reduced flexural requirements, the center span has less post-
tensioning than the exterior spans. The support joint therefore opened on the interior side
of the interior pier segment. The final deflected shape of the structure clearly illustrates
the mechanism behavior of the structure at ultimate load levels.
Because the external tendons are bonded to the concrete section only at discrete
locations along the span, large concentrated rotations must occur at opening joints to
develop the large tendon elongations required for increased tendon stresses. These rotations
allow the internal forces to redistribute to stiffer uncracked regions. This is apparent from
the reaction and joint-moment data for the flexural test of the north span, shown in Fig. 11.
If no redistribution had taken place, these plots would be linear. The changes in slope
indicate redistribution of moments and the related reaction is occurring. The initial slopes
reflect the expected behavior from an elastic analysis. As the midspan joints begin to
open at the decompression load, the resultant loss in stiffness causes a larger portion of
 Prestressing 321

the additional load to be carried at the interior support. As loading is further increased,
the support joint opens causing a reduction in stiffness at the support as reflected by the
increased reaction rate. The internal forces then redistribute back towards the midspan
region with the ultimate distribution of internal forces being controlled by the relative
stiffness of the support and midspan regions.
This redistribution of internal forces, caused by "hinging" at the critical joint
locations, will also cause redistribution of the secondary prestress forces near ultimate
load levels. The secondary prestress forces are caused by geometric constraints on the
entire structure when the tendons are initially stressed (2). To develop the required tensile
forces with external tendons, large rotations must occur at the segment joints. As the
joints "hinge" and the mechanism forms, the forces from the initial geometric constraints
dissipate. If the segments are detailed to allow large rotations to occur at the segment
joints, then the geometric constraints will no longer be valid. The geometric constraints,
and the corresponding secondary forces therefore affect the service load behavior, with the
ultimate condition approaching the plastic mechanism behavior.
As loads are increased beyond service levels, the tendon stresses exhibit several
stages of behavior, shown in Fig. 12a. The concrete stress profile at the critical opening joint
is shown in Fig. 12b for important stages of tendon stress development. Initially, before
the joints begin to opep., the tendon-stress increases are linear with the applied load. The
tendon stresses remain linear until the neutral axis at the opening joint reaches the level
of the tendon, Point B, at an applied load that is slightly greater than the decompression
load, Pd. Beyond this load, the tendon stresses increase slowly at first as the increased
moments are resisted primarily by an increased internal-force lever arm. When the resultant
concrete compressive stresses are concentrated in the top flange of the section, Point C,
then additional moments must be resisted by increased tendon forces. To develop the
required tensile forces with external tendons, large rotations must occur at opening joints
resulting in increased deflections and joint openings.
The applied-load stresses for a typical tendon during the flexural strength load
cycle are shown in Fig. 13. Tendon strain measurements were made at the exterior and in-
terior ends of the span, joints (1,2) and (9,10) respectively, and at midspan, joint (5,6). The
midspan tendon stresses remained linear with applied load up to approximately 1.8*(11+1)
when the concrete section had decompressed to the level of the external tendons. This load
is slightly higher than the measured decompression load of 1.4*(11+1). The tendon stresses
increased slowly at first until the midspan joints opened at 3.0*(11+1). At this load level
the resultant compressive stress had concentrated in the top flange and additional mo-
ments were resisted by a direct increase in tendon stress. Subsequently, as the support
joint opened, at approximately 4.8*(11+1), midspan moments increased and the rate of
tendon stress development also increased.
At an applied load of 5.0*(11+1) the tendon began to slip from the interior end
towards the midspan region. Slip also occurred from the exterior end at approximately
6.2*(11+1). The tendon slipped through the deviator when the change in tendon force
exceeded the maximum friction capacity. Substantial tendon slip was noticed in all tendons
at all locations for ultimate load levels.
The increase in midspan tendon stresses corresponding with flexural strength
ranged from 35 to 60 ksi in the midspan region and from 15 to 20 ksi at the critical
support joint. The ACl formula for unbonded post-tensioning tendons (3) was used to
predict the ultimate tendon stresses at midspan and at the critical support joint. The
322 MacGregor, Kreger, and Breen

average measured tendon stress increases in the midspan regions were accurately predicted
by the ACI formula with measured-to-predicted ratios of 0.98 in the north dry span and
1.03 in the south epoxied span. This result is reasonable if it is remembered that the
tests used to develop the ACI formula {5) were conducted on specimens with short span
lengths approximately equal to that used in the model bridge. The AASHTO formula
{4) for unbonded tendons predicts lower stress increases, but implicitly assumes much
longer spans. The slightly higher stress increases in the epoxied span may be due to
the concentration of rotation at a single joint causing larger induced deformations in the
tendon.
The average measured tendon stress increase at the critical support joint was
overestimated by the ACI formula with measured to predicted ratios of 0.61 at the north
interior support and 0.69 at the south interior support. At these locations the effective
depth of the external tendons is reduced because of the drape from the support. An increase
in the ratio of the free length of tendon segments to the effective depth of tendons leads to
reduced stress development under applied loads.
The maximum midspan stress that was achieved in the model tendons was af-
fected by the load level at which slip began. If tendon slip began at a low load level then
the ultimate midspan tendon stress was low. Conversely, if slip did not occur until higher
load levels then the ultimate midspan tendon stress was increased. Therefore, before proto-
type extrapolation can be made, additional information is required to determine the force
transfer mechanism at deviators.
A final comment can be made about the overall performance of the structural
system. The maximum applied-load moment at the center of the north span, for an applied-
load equal to the design service live load (DL+(LL+I)), was approximately 50 ft-kips.
When the ultimate flexural strength was reached in the north span, at an applied load
of 1.3*DL+6.8*{LL+I), the maximum applied-load moment at the same midspan location
was approximately 250ft-kips. With the additional factored dead load, 0.3*{DL), causing a
midspan moment of approximately 30 ft- kips, the overall factor of safety above the service
load condition, with respect to the midspan moment, is approximately (250+30)/50 or 5.6.

Joint Behavior

The local behavior of the segments near an opening joint was affected by the
amount of shear that was being transferred across the joints. In the flexural tests, with the
load applied as a series of forces along the longitudinal axis of the structure, small shears
were transferred across the critical opening joints. In this case the concentrated rotations
occurred either at the joints in the dry span or at a crack adjacent to the precast joint in
the epoxied spans. At ultimate load levels the joint/crack had opened into the top flange
of the girder in both the dry-jointed and epoxy-jointed spans.
The local force transfer mechanism in the segments adjacent to the opening joints
when the flexural strength was reached is shown schematically in Fig. 14a. The joint/crack
had opened into the top flange causing the load to arch across the segment joint. The small
shears that are transferred across the open joints at this stage are carried by the vertical
component of the "arch force" at t!Je joint. The segment reinforcement transfers the shears
from the load point to the edge of the segment, and then the arch action transfers the force
across the joint.
In the shear tests, a concentrated force was applied to the structure so that
significant shear would be transferred across opening joints. In this case, after the joint
 Prestressing 323

had opened up through the bottom flange, an inclined crack formed from the load point
to the bottom of the web at the edge of the segment, as shown in Fig. 14b. As load
was increased to ultimate levels, the concentrated rotations occurred at the inclined crack
leaving the joint region in firm contact. This was true for both the dry-jointed and the
epoxy-jointed spans.
The local force transfer mechanism in the segments adjacent to the opening joints
when capacity was reached is shown schematically in Fig. 14b. A compressive strut formed
from the load point to the lower corner of the segment. The segment web reinforcement
transmitted this force across the inclined crack to the top of the segment. The shear force
was then transferred across the joint utilizing much of the depth of the webs.
The reinforcement for the concrete segments near opening joints must be properly
detailed to allow the large rotations required for tendon stress increases. Local truss mech-
anisms, such as shown in Fig. 14, should be developed for the critical segments to ensure
that force transfer can be made across the joints. The bottom longitudinal reinforcement
must be anchored close to the opening joint and must resist the horizontal component from
the transient shears (Fig. 14c) plus an inclined strut running from the load point to the
bottom corner of the segment. The web reinforcement must be able to resist the transient
shear from global loads plus the vertical component of the inclined strut. The web rein-
forcement must be anchored under the bottom longitudinal reinforcement and high in the
section so that anchorage is maintained when the neutral axis shifts to the top flange of
the segment.

CONCLUSIONS

Several important observations have been made concerning the full range of be-
havior of segmental box-girder bridges with external post- tensioning tendons.
1. The structure was extremely stiff for service load conditions, with live-load de-
flections of about L/6000 in the exterior spans and L/7500 for the interior span.
Tendon slip was not noticed during service load cycles. Additional information
is required to assess the problem of fretting fatigue at the deviators of external
tendons.
2. The cracking load in the epox:, joints was approximately twice the load required
to decompress the flexural tension fiber and begin to open a previously cracked
joint. Cracking occurred through the concrete adjacent to the epoxied joint.
3. The structure exhibited linear behavior to load levels higher than the factored
design ultimate load.
4. The maximum applied-load moment in the north span, when flexural capacity
was reached, was approximately 5.6 times the maximum applied service-load
moment in the north span. This indicates that the applied midspan moment has
an overall factor of safety above the service load condition of approximately 5.6.
5. Large concentrated rotations are required at opening joints to cause tendon
stresses to increase with the applied load. These rotations allow the internal
forces to redistribute to stiffer regions. The secondary prestress forces also redis-
tribute as ultimate load levels are reached.
6. The applied-load tendon stresses at midspan of the loaded span, corresponding to
the flexural capacity of the girder, were accurately predicted by the ACI formula
324 MacGregor, Kreger, and Breen

for unhanded prestressing tendons. The AASHTO formula for unhanded ten-
dons underestimated the ultimate tendon stresses at midspan. The applied-load
tendon stresses at the critical support joint, corresponding to the flexural capac-
ity of the girder, were overestimated by the ACI formula and underestimated
by the AASHTO formula. Additional information is necessary, however, before
extrapolations can be made to the prototype structure.
7. Tendon slip was noticed at the deviators in all the tendons at all locations during
the ultimate strength cycles. The tendons also slipped during cracking and joint-
opening cycles.
8. The local transfer of forces across opening joints depended on the level of shear
being transmitted across the joint. For opening joints with small shear transfer,
the joint/crack opened in a flexural mode into the top flange of the structure
with the concentrated rotations occurring at the joint. For opening joints with
large shear transfer, an inclined crack formed from the load point to the lower
corner of the segment adjacent to the joint. The concentrated rotations occurred
at the inclined crack.
A more complete discussion of the test setup, procedures, observations, and con-
clusions are available in Reference 6.

ACKNOWLEDGEMENTS

This paper is based on research sponsored by the Texas State Department of

Highways and Public Transportation and the Federal Highway Administration. In addition,
substantial financial support for the research program was provided by the National Science
Foundation under Grant ECE-8419430. All opinions and conclusions expressed are those
of the authors and do not necessarily represent the views of the sponsors.
The authors would like to particularly acknowledge the contributions from Mr.
Alan Matejowsky of the TSDHPT who provided valuable suggestions and practical insight
throughout all phases of the research project. In addition, the authors would like to
acknowledge the guidance and assistance provided by local industry in development and
construction of the model bridge structure. In particular, the assistance and cooperation
of Pres con Corporation of San Antonio and Ivy Wire and Steel of Houston were especially
appreciated. Finally, the hard work and personal contributions by Mr. Elie Homsi of
Prescon Corporation were greatly appreciated.
 Prestressing 325

REFERENCES

1. Yates, "A Study of Fretting Fatigue in Post-Tensioned Concrete Beams" , unpub-

lished M.S. Thesis, The University of Texas at Austin, December 1987.
2. Lin T .Y., and Thornton K. , "Secondary Moment and Moment Redistribution in
Continuous Prestressed Concrete Beams", PCI Journal, Jan-Feb 1972, Vol. 17
No. 1.
3. American Concrete Institute Committee 318, "Building Code Requirements for
Reinforced Concrete (ACI 318-83), American Concrete Institute, Detroit, Michi-
gan .
4. AASHTO, "Standard Specification for Highway Bridges", 13th Edition, Ameri-
can Association of State Highway and Transportation Officials, 1983.
5. Mattock, A.H., Yamazaki , J ., and Kattula, B.T. , "Comparative Study of Pre-
stressed Concrete Beams, With and Without Bond" , ACI Journal, Proceedings
Vol. 68, No.2, Feb . 1971, pp 116-125.
6. MacGregor, R.J .G., "Strength and Ductility of Externally Post-Tensioned Seg-
mental Box-Girder Bridges" , PhD. Dissertation, The University of Texas at
Austin, December, 1988.

FIG.1 - BOX-GIRDER BRIDGE MODEL

 CJ,;I
SYMMETRICAL _:_ t\J
~~~ 0\
I
r-PIER SEGMENT rTVPICAL SPAN SEGMENT C.I.P. CLOSURE STRIP
}
I
s=
~
l.l
••••••
!I I: :I I: :I !I I: :I I: II I: :I !I I: :I I:

---- ------- ------ ------- ------ ------ ------- ------ ------

~
------ - ------ ------- ------ ------ ------- ------· ~
:I II: !I I: :I :j l! Lt J: II J: :_l :_1 J! ! I I: ~""!

± I
~
(:6
7'-0" TOP FLANGE I
(Jq
('!I
4'-0" BOTTOM FLANGE
PLAN VIEW ~""!

I
§
r16" I 0...
~
(:6
('!I
::I

3" 3"
10
77'-0"

ELEVATION

FIG.2 - MODEL DIMENSIONS

 Prestressing 327

7'-0"

SECTION SIDE VIEW

NOTES:
SPAN to DEPTH RATIO = 18.75
EFFICIENCY FACTOR .60 =

a. TYPICAL SPAN SEGMENT

7'-0"

.................. ... ~
~
II"[
__ c,.cocc,. ...
\
16-1"_ _
•
3.5___
~0
c,.,.",."'c.-.c"",.,..l\."'..,"'..,"'~

:.:.>>:.:.:.:
•••••••
• ••• •. •. •. •. •. •
&0
EXTERNAL
TENDONS
............................ "" ......... ,. ...
.-o. I· ANCHORAGE
I
·
4
REGION

SECTION SIDE VIEW

b. PIER SEGMENT

FIG.3 - MODEL CROSS-SECTIONS

328 MacGregor, Kreger, and Breen

SEGMENT END ELEVATION

1-;--1'
"!

M '<t
M

l_s_
0

2~2 "
" •

"

5"
r~

WEB DETAIL SHEAR KEY DIMENSIONS

FIG.4 - SHEAR KEY DETAILS

Fti4:cn-rrr.P.f . _ 1'-~lJ-~tl~1-~1-~-t-n-1].;:.1:::
~:~:t::::t::·::J:::t::±::±:::r:::t:5:':: t::±:::t=§:l:~l~~4:~~~~:J~f?:¥:E:b~:::

·-.-,- r -.-.-r.-.-,T-.-r;,·.r:·r~;.r-~--r.-~-~;.-r;;.-r.--.;.-r-;;;.-;; r:;;J:.-;.r;--.r.:ot;,·-1u---iE~1-FR"-T.-Fi-Tr-;£:;£it·-r;----

m~il:IDmBmrt!B§=JJ§E![fTI:§:gmrmmm~tflL ~-PITit:t=
IE 6 I 7 I a I I I 1 o I HI 11 I 12 I 13 I 14 I u I 11 111 111 I 11 I 20 II Sli 2112 2 I 2 3 I 24 I 2 s I 211 27 I 211 21 I 30 I se
NORTH ""---SEGMENT HUMBER SOUTH
PLAN VIEW

ELEVATION
3 1
; : } AT NOIITH UT. a'-"'POftf IUCTION AI )
5,S,Ie.l,1A AT NOftTH ..r. SUPf'OfU {SECnoH Cl)

~
'-.!·}
.,~,, ..
AT SOUTH llfT. IUP'I"'RT (SE:CTIOH A2)
~-;·:,.·;
rJl
......
"1
('!I
rJl
SECTION A SECTION 8 SECTION C rJl
s·
O'Cl

FIG.5 - SCHEMATIC POST-TENSIONING LAYOUT CJ.i

t-.:1
1.0
330 MacGregor, Kreger, and Breen

'w
CfJ

--
I t)
N
-
I t)
N

CfJ

CfJ
rn t-
c. t-
w CfJ
..:.::
~
,...
N --
I t)
t-
..J
c(
a:
rn
c.
..:.::
--
It)
w
t-
a:
c(
c
<t
0
..J

-+
N N
II ::;)
~ w I
1-
><
w ,...
0 :I: C/)
w
..J
..J
""":"
..J
LL
-II
+
..J
1-

co

-
..J (!)
u:::

a.
a. a.
N
--
I t)
N
 DEFLECTION PROFILE
Service Load Tests of North Span - Live Load Cycles
.03~----------------------~----~------~--~

.01

_..
CD
Q)
.r:. -.01
0
c
~

c
0
:;:o -.OJ
0
Q)
;;:
p p
Q)
0 f_11_,~tt_f !!l. 2P:EQU.(LL+I):12.4 kips

-.05 ~~~~~~~r1-r 1 1 ~ 1 1 1 1 r 1 r 1 1 ft r 1 1 1 1 1 1 1 1 A
-~511 I 2511 I 2511 i i
[Jl
.......
~
-.07~--------------~--------------~--------------~
[Jl
[Jl

0 25 50 75 s·
(J'q
Location (ft from N.E.)
FIG.7 - TYPICAL SERVICE LOAD DEFLECTIONS w
w
.......
332 MacGregor, Kreger, and Breen

DEVIATOR SCHEMATIC 12 >II

Vs, SURFACE SHEAR

APPLIED TO WIRES

FIG.8 - DEVIATOR SCHEMATIC

 APPLIED LOAD vs DEFLECTION
Flexural Strength Tests of North Span - Ultimate Cycle
8~--------------------------~----------~--~

7 SUPPORT JOINT OPENS

+
__J
__J
-MECHANISM FORMS
8
......
c :s.JPm
Q)
ctl
~
:::J 4
0" MIDSPAN JOINTS OPEN
w I
"'0 J
ctl p p
0 2P:EQU .(LL+I)=12.4 kips
__J 2
"'0
Q)

Q_ 51. SL
~
........ 25tt 1 25tt
0...
<( 1.3* DL (/)
......
~
(/)
-1 +---------~---------r--------~--------~~--------T---------,---------~--------~--------~ (/)

0 .2 .4 .8 .e 1 1.2 1.4 1.8 1.8 s·

CfCI
Deflection (inches)
FIG.9 - APPLIED-LOAD vs DEFLECTION w
w
w
 eN
DEFLECTION PROFILE eN
~
Flexural Strength Tests of North Span - Ultimate Cycle
"
4
I 1.3* DL+{ LL+I) c::: 1. 3* D L+2.86*{L
•
L+l) I
~==
C1
(:6
(Jq
0
~..,

............ :;:::
01 -.4 (:6
CD (Jq
.s::. 1. 3*DL+3.0*(LL+I) ('II
(J ~..,

c: MIDSPAN JOINTS OPEN

.:.::,. ~

c:
!;j
0
-.8

1.3*DL+4.8*{LL+I)
I ::l
0...
ttl
(J (:6
CD -1.2 SUPPORT JOINT OPENS ('II
;;: ::l
CD
Cl
I
-1.6 .f v ""'- 1.3*DL+6.8*{LL+I)
FLEXURAL STRENGTH

-2
I I
0 2S so 7S
Location (ft from N.E.)

FIG.10 - ULTIMATE LOAD DEFLECTIONS

 Prestressing 335

REACTIONS vs APPLIED LOAD

Flexural Strength Test of North Span - Ultimate Cycle
100~----~~--~------------~------------~--~

..
......
:X
a. 50
......
..
c
0
NE
!fl0 25
•
a:: SE

0 ---
-25+----+----~---,----~--~r---~----r----r--~
-1 1.31>0L 2 3 4 5 7
• a
Applied Load - Equivalent (Ll+l)

JOINT MOMENT vs APPLIED LOAD

Flexurol Strength Test of North Span - Ultimate Cycle
400~--------~~D---a-------~----------~-----,

(4,5)
200

...... 100
~
I
s...
e
c

0
-100

:2 -200 (NI,11)

-300

-400+----+----~--~----~--~----~--~----~--~
-1 1.3• DL 2 3 4 5
•
Applied Load - Equivalent (LL+I)
7
•

FIG.11 - REACTIONS AND JOINT-MOMENTS

 w
w
TENDON 0"1
STRESS
D
~

{ il t [.I
~
A B
c ~
rse 1 e ....., 0
(:6
O'Cl
0
~'"!
A INITIAL Pd- DECOMPRESSION
CONDITION LOI\11 ~
(:6

~
Pd Pm

APPLIED LOAD

D
tz:.r t k .r
B. TENDON STRESS
BECOMES NON-LINEAR
C. COMPRESS! VE STRESSES
CONCENTRATE IN TOP Fl ANGE
~
~'"~

~
=
Q...
t;l:j
(:6
Cb
=

t I .1 t I .1
L>.P
c

A B
Pd Pm Pu
Pm- MECHANISM LOAD DULTIMATE
APPLIED LOAD CONDITION

a. TENDON STRESS RESPONSE b. KEY PO I Nrs FOR TENDON STRESSES

FIG.12 - TENDON STRESS RESPONSE TO APPLIED-LOAD

 CHANGE IN TENDON STRESS vs APPLIED LOAD
Flexural Strength Tests of North Span - Tendon 1 b - Ultimate Cycle
SOT-----~------------~------------------~

50

~
[fJ
.....
@
-10+---~~--~-+--n-----r----,----+r----~---,.---~ [fJ
[fJ
-1 1.3* DL 1 2 J 4 S 8 7 8 s·
~
Applied Load (Equivalent Model Uve Loads)

FIG.13 - TYPICAL TENDON STRESS RESPONSE w

w
-.J
338 MacGregor, Kreger, and Breen

p
V 1 > V2
V1+V2=P

EXTERNAL _ __.
a. FLEXURAL TEST TENDON

2P V1 = V2
V 1+V2=2P

b. SHEAR TEST

c. GENERAL

FIG.14 - SHEAR MECHANISMS AT OPENING JOINTS

 SP 120-16

A New Methodology for

the Analysis of Beams Prestressed
with External or Unbonded Tendons
by A.E. Naaman

Synopsis: A simple methodology for the solution of beams prestressed or

partially prestressed with external or unbonded tendons in the linear
elastic cracked and uncracked range of behavior is described. It leads to
equations allowing the computation of stresses in the concrete section, the
tensile reinforcing steel, the compression reinforcing steel, and the
prestressing steel. In particular, it is shown that the stress in unbonded
tendons is a function of the applied loading, the steel profile, and the ratio
of the crack width (or crack band width) to the span. These factors can
all be accounted for through the use of a strain reduction coefficient n
for the uncracked range of behavior and a similar coefficient nc for the
cracked range of behavior. It is shown that, when the strain reduction
coefficients n and nc are taken equal to unity, the solutions developed
here revert to the solutions developed earlier for partially prestressed
beams with bonded tendons.

Keywords: beams (supports); cracking (fracturing); flexural

strength; partial prestressing; prestressed concrete;
prestressing; serviceability; structural analysis; structural
design; unbonded prestressing

339
340 Naaman

Fellow ACI Antoine E. Naaman is a Professor of Civil Engineering at the

University of Michigan, Ann Arbor. He obtained his Ph.D. degree from the
Massachusetts Institute of Technology in 1972. He is active on several
ACI Committees including Prestressed Concrete, Concrete Bridges, High
Strength Concrete, Fiber Concrete, and Ferrocement. His current
research interests include partially prestressed concrete, external
prestressing, and high performance fiber reinforced cement composites.

INTRODUCTION

External prestressing implies the use of prestressing tendons

outside the concrete section of a structural concrete member. These
tendons are primarily designed to ensure horizontal prestressing of the
structure and generally represent only a _portion of its total
reinforcemel)t. The remaining reinforcement may consist of reinforcing
steel, prestressing steel, or a combination of them. External
prestressing is being increasingly considered in the construction of new
concrete structures, particularly bridges (1 ,2), and is a primary
method for the rehabilitation and strengthening of old structures (3).

Externally prestressed tendons are not bonded to the concrete.

Thus, from an analysis and design viewpoint, the)'_can. be. treated as
"un1:>o11ded tendons", assuming that secondary effects are negligible. The
use of unbonded tendons may not only imply ungrouted steel tendons, but
may also refer to the use of unbonded polymeric tendons. Therefore from
the analysis, design and constructional aspects, external prestressing,
unbonded tendons and polymeric tendons are part of the same family of
problems and could oeaadressed-by the same general solution.

The use of external prestressing. to rehabilitate a reinforced

concrete bridg-e leads to a structural system referred to as partial
pr~~!r_~~§ing. Although recent years have se-eri -extens-ive-advances inthe
analysis and design of partially prestressed members(4 to 7), little
information exists on how to accommodate in the analysis unbonded or
external tendons .

As the use of external or unbonded tendons poses some uncommon

problems in analysis or design, rational methods must be developed to
deal with such problems. Rational analytic solutions would allow the
accurate determination of stresses and strains at any section along the
member and verification of serviceability and strength limit states such
as cracking, fatigue, long term deflections and the like. Recognizing the
unusual nature of the behavior of beams prestressed with unbonded
tendons, several studies have attempted in the past to provide
clarifications to such behavior and to address various aspects of their
analysis and design under service and ultimate loads ( ~ to 18).

The main objective of this paper is to present a rational,

simplified methodology for the analysis of beams prestressed with
external or unbonded tendons throughout their linear elastic cracked and
uncracked range of behavior. Particular_ emph(lsis is giv(3_1l __ to
determining stresses and strains in the constituent materials, steel and
 Prestressing 341

concrete, for commonly applied loadings and tendon profiles. It is pointed

out that the main feature o_f_!~e pro_pos_ed method is to reduce the analysis
of beams with unbonded tendons to that of beams with bonded tendons
throiJQ~:::_t~:e_=u~e~¥Sfmple predeteri_T_ll~ecfcoettiCTents-:---·--·

FLEXURAL ANALYSIS OF BEAMS

A schematic representation of the moment deflection relationship

of a beam prestressed or partially prestressed with external or unbonded
tendons is shown in Fig. 1. This curve could a).S.9 represent tlla.moment
curvature relation of a beam sec:!ion. The curve can be conceptually
divided into several parts: part AB corresponds to the linear elastic
uncracked state of behavior, part BC illustrates the onset of cracking,
part CD represents the linear elastic cracked range, part DE corresponds
to the cracked nonlinear range of behavior, point E represents the
nominal or ultimate resistance, and point F the failure point. The curve
of Fig. 1 describes the range of possible behavior. In theory a s.udden
change in deflection occurs at the onset of crackirig--(segment. .B.C);
however the real behavior shows a more gradual change in slope Jr.om_A!3
to DE.

Actual design of beams with external or unbonded tendons may or

may not allow cracking under service loads. However, as loads are
random in nature and overloads are quite common, t~ _cr_acking state
should be considered in any comprehensive evaluation. Moreover,
cracking is allowed by the ACI Cgde Jn partially prestressed members
usjng bonded tendons, and Tri-slabs with unbonded tendons provided
sufficient amount of non-prestressed conventional steel is added. In U.S.
practice, both service load design (i.e. working stress design or allowable
stress design) and ultimate strength design are required for prestressed
concrete structures. Service load design implies linear elastic behavior
in the uncracked and cracked range. ~-~.!!!!l.!!l~!LQ!_!he. an~l~is
procedure suggested here for the line_~!..~Jastig_!~_~_e~_r~_:;~-~9iv~-"-- below.

Linear Elastjc Analysis jn the Uncracked State

This state is represented by segment AB in Fig. 1. The material

components, steel and concrete, are assumed to work in their linear
elastic range of behavior. However, no bond exists between prestressing
tendons and concrete.

It is commonly assumed that, in this range of behavior, the change

in the prestressing force F d_u_e_t2 aR2[L~d _IQadj_!!g__ ~~- n_E)gligible. Thus F is
a§sumed to remain constant anc.lthe analysis of the section for stresses anct
strains
- is similar to that with bonded tendons.
. . . . ·-·--

If for the purpose of exact analysis or accuracy, the change in the

prestressing force with applied loading is needed, then the analysis of a
prestressed beam under load can be much more involved than that
assuming constant F particularly if the tendons are unbonded. For the
case of bonded tendons a complete solution assuming a variable value of F
has been already described in an earlier publication (19). For unbonded
tendons the methodology described below, consists of reducing the
342 Naaman

problem to the case of bonded tendons through the use of a strain reduction
coefficient n. It can be shown that the coefficient n depends only on the
steel profile _and txee of lol:l.ding, and needs to be detE;~rmined.o!lly once.Jor
comfif6iiloading-ancftendon
-~-
configurations.
----"-· ·--------·· ______, _____
~-

The following assumptions are made for the analysis:

a. the materials are linear elastic _Ln. th~U~n.g_e_9f b~havior considered,
b. linearsrratn-distriblition is assumed along the concrete section, and
c. second order effects, if any, for external tendons are negligible.

It ~~-~ssume_d_ that the critic~l design section is the midspan

section at which the maxiinum-ecceritrlcityoT tne- tendons is assigned.
Simply supported beams are considered first.

The following equations provide a methodology to determine the

stress change in the prestressing steel for any moment larger than the
dead load moment, Mo. and smaller than the cracking moment, Mer.

Mo < M < Mer (segment AB of Fig. 1)

The reference state is defined as the state-~l_oadJD.g__@rresponding

to Jhe combined action of the effective prestressing force (after losses)
a_n~U!!E:!_ c:l~~d_lgad moment . The corresponding stress diagram along the
section is shown in Fig. 3~ Thus:

ForM= Mo I J <) ; ·J ' /

fps = fpe (1)

( 2)

where fps is the stress in the prestressing steel at any loading state, fpe
is the effective stress in the prestressing steel, Fe is the effective
prestressing force, and Aps is the cross sectional area of the steel.
!o--,, J
For a moment larger than the dead load moment and smaller that
the cracking moment, the following equation can be written (Fig. 3):

( 3)

where Mps represents a 9hange in stress in the prestressing tendons.

Forbondedtendons:
~----- - -· - ~

(Mps )bonded = Eps (~Eps}bonded (4)

where ~Eps represent the strain change in the prestressing steel in the
section considered due to an increment in bending moment ~M = (M-Mo).
It is also equal to the strain change in the concrete at the level of the steel
afthe section considered, that is:

(~Eps)bonded = (~Ecps)bonded ( 5)
 Prestressing 343

If the section of maximum bending moment is being analyzed, then

~ec_p~ is also the ~_!r_alll_~~ange in_JI!~- criJic::_Cl_l_(l~~-~~~~~?~ (here the
midsp~~-~~~t~?n), (~Ecps)~.

For external or unbonded tendons the change in stress in the

tendons is assumed given by:

(~eps)unbonded = (~eps)average (6)

where (~Eps)average is .ttJe._a'iefage strain increase in the external tend~n

over the length of the member. Let us define the strain reduction
coefficient n as:

(~epslm unbonded (~Fj'osl average

( 7)
(~epslm bonded (~Ecpslm bonded

where the subscript m implies the midspan section or the s~~~?_r_:_of

maximum moment.

Note that for bonded tendons, n equals 1. For unbonded tendons n

can be calculated in the most general form from thetOTIOwing equatior;:-

2 rv2 M(x) e0 (x) dx ( 8)

Mmax (eolmax I J0
Applying the above equation to the loading cases and profiles shown
in Fig. 2 leads to the values of n listed in Table 1.

The use of the coefficient n can be integrated in the analysts of

the uncracked section to generate expressions for the stresses and strains
in !!1~L~teel and concrete at the critical section (i.e. section of maximum
moment and maximum eccentricity). The solution equations are given in
Table 2 and apply to any state of loading between the reference state and
the cracking state (Fig. 3) assuming linear elastic uncracked section
behavior. It can be observed that for n = 1, these equations revert to the
equations given in (19) for bonded tendons.

Linear Elastic Analysis in the Cracked State

This part of the analysis covers segment CD of the moment

deflection curve described in Fig. 1.

If the applied moment exceeds the cracking moment, a crack will

appear in the midspan region. For the purpose of analysis, a§§Ur:!l~Jh._at
only one crack will form at the section of maximum _rT1011'l~nt linc;U~L\,1?
analyze the beam assuming that it is dividedlnto-iwo-parts, one uncracked
portion with moment of inertia lg and one c_r~~~-~~__e<>_riLi?.!LV.V1i!fmQ_fn~e_nfof
in9"r_fia-1cr (f:ig. 4). The cracked portion has a length lc while the
uncracked portion has a length (He). In a manner similar to what was
344 Naaman

done in the previous section, Ol!_~al')_ define a bond coefficient nc which

represents the ratio of average strain change in the unbonded tendon to tiie
strain change in the concrete at the level of the steel at the sectiQn of
maxim urn moment. This last i~_rn is also equaUQ Jll!Lslr~Jo_~haoge in__1~_e
equivalent bOnded tendooassuming a craCked section. It can be shown
that, for symmetricai-ioading and tendon· prome;1fie value of nc can be
obtained in the most general manner from the following equation:

nC =
nlcr
-
lg
+ -2 (1 - ~
1
lc r
lg I
Jic/2 M(x) e 0 (x) dx
( 9)
0
Mmax (eo)max

where Mmax and (e 0 )max represent the applied external moment and the
eccentricity of the tendons at the section of maximum moment (here the
midspan section).

The value of nc has been calculated for the various combinations of

loadings and tendon profiles described in Fig. 2. The solutions obtained
are summarized in Table 3. It can be observed that the value of nc
depends on:
1) the value of n for the uncracked state,
2) the ratio of cracked moment of inertia at the section of maximum
moment to the gross moment of inertia, and
3) the ratio of length of the cracked zone to the span length.

To illustrate how nc varies with these variable, a plot of nc for a

typical loading and tendon profile is shown in Fig. 5 . It can be observed
that, while n c is greatly influenced by the ratio lcrllg, it is not as
sensitive to the value of Idl at small values of lc/1. Indeed one can assume
that as a first approximation, the value of nc is equal to the value of
nlcrllg. Going back to Table 2 where the values of~are~n_, it can
be_g~~..ed. thaLO-G_£an __l?~ approximated in all cases by Olcrflg provided
.lc/1 is~. This is indeedthe case,-since T~card:ieseen-as-tflewidth of
the crack at midspan or the width of a plasticized region with smeared
In
cracking Uie midspan region. Thus, for practicar cre-sign pUrposes:-

( 1 0)

Th!')__mathematical equations expressing _ equilibrium,

compatibilit}'_. and _stress-strain relationships assuming that the
CQilstit!,J~!llmaterials remain .in the. Ji11ear e'C!Stic Jang,e_ot.beha'@r~ ~n
bg,~~isirniJarly to the case oLt_I1JL@al~o!Jl..9J~9JSe.9_~~cljQJl_Gol])aining
bonde_~ pre_stressJng steel and conventional reinf()rcing__bars. T~rain
compatibility equations will contain the c_()effLc[e_l!_l_'l.c· These equations
canthen.be-sorved: The general solution equations are given in -Appendi?< I
with -some notation explained in Fig. 6. The figure shows the strain
diagrams along the section for the reference state and the current state of
 Prestressing 345

loading. Alsg shown are the corresponding strain in the prestressing

steel.

The first equation of Appendix I is a cubic equation in c where c is

the depth to the neutral axis (zero stress point) of the elastic cracked
section. As nc must be estimated from Eq. 10, and as lcrllg depends on c,
some iteration in the solution of Equation 1 may be needed. The following
computational steps can be followed: assume a value of lcrllg , determine
c from Equation 1, compute corresponding lcr· and check if the assumed
lcrllg is acceptable. Otherwise, iterate. Once an acceptable value of c is
obtained, the corresponding stresses in the constituent materials, steel
and concrete, can be determined from the equations of Appendix I. In
particular the stress in the unbonded prestressing steel can be calculated
for any applied moment larger than the cracking moment ans smaller than
the moment leading to inelastic behavior of one of the components
materials. Note that the solution equations given in Appendix I cover
prestressed and partially prestressed rectangular and T
sections with and without compressive reinforcement. When the value of
nc is taken equal to 1, these equations revert to the case of beams
prestressed with bonded tendons and become identical to the solution
equations derived earlier in (19).

CONCWDING REMARKS

The methodology described in this paper allows for a rational yet

simplified analysis of beams prestressed with external or unbonded
tendons in the linear elastic cracked or uncracked range of behavior. It
provides a link between the case of bonded tendons and the case of
unbonded tendons through the use of strain reduction coefficients n and nc
which can be computed only once for common combinations of loadings and
tendon profiles. Although the solution described abo_ve allow Jor the
analysis of several serviceability limit states, it is conceivable that the
concept of strain reduction coefficient o can be extended to cover the
ultimate flexural strength limit state (20). However, substantial
additional research is needed to ascertain the validity of application of
such methodology to the ultimate strength limit state.

ACKNOWLEDGEMENTS

The research work of the author in the field of prestressed concrete

has been supported in part by Grant No. MSM 86-13440 from the
National Science Foundation to the University of Michigan, with Dr. G.
Albright as program director. Their support is gratefully acknowledged.

REFERENCES

1. Virlogeux, M., "La Precontrainte Exterieure," Annales de 1'1TBTP,

Paris, 1983, pp. 115-191.

2. Virlogeux, M., "External Prestressing," IABSE Proceedings,

P62/83, pp. 101-108.
346 Naaman

3. Nowak, A.S. and Absi, E., Editors Proceedings of the first US-
Eurpoean Workshop on "Bridge Evaluation, Repair and
Rehabilitation," Paris, June 1987, University of Michigan, 705
pages.

4. Proceedings of the FIP Symposium on Partial Prestressing and

Practical Construction in Prestressed and Reinforced Concrete,
Bucarest, Romania, September 1980, Vol. 1-2.

5. Proceedings of the Conference on Partially Prestressed Concrete

Structures, Bratislava, Czechoslovakia, June 1988, 98 pages.

6. Naaman, A.E., "Partially Prestressed Concrete: Review and

Recommendations," Journal of the Prestressed Concre1e
Institute, Vol. 30, No. 6, November-December 1985, pp. 30-71.

7. Cohn, M.Z., Editor, "Partial Prestressing: From Theory to

Practice," Vol. 1 and 2, Martinus Nejhoff Publishers, Dordrecht,
The Netherlands, 1986.

ACI-ASCE Committee 423, "Recommendations for Concrete

Members Prestressed With Unbonded Tendons," American
Concrete Institute, ACI 423.3R-83, 1983.

9. Balaguru, P.N., "Design of Partially Prestressed Post-Tensioned

Beams With Unbonded Tendons", Concrete International, Vol. 3,
No. 11, November 1981, pp. 30-41.

1 0. Balaguru, P.N., "Increase of Stress in Unbonded Tendons in

Prestressed Concrete Beams and Slabs", Canadjan Journal of Cjyi!
Engineers, Vol. 8, January 1981.

11. Burns, Ned H., Charney, F.A., and Vines, W.R., "Tests of One-Way
Post-Tensioned Slabs With Unbonded Tendons", PCI Journal,
September-October 1978, pp. 66-83.

1 2. Cooke, N., Park, R., and Yong,P ., "Flexural Strength of Prestressed

Concrete Members With Unbonded Tendons", PCI Journal, Vol. 26,
No. 6, November-December 1981, pp. 52-80.

13. Elzanaty, A., and Nilson, A., "Flexural Behavior of Unbonded Post-
Tensioned Partially Prestressed Concrete Beams", Research Report
No. 82-15, Cornell University, November 1982.

14. Mattock, A. H., Yamazaki, J., and Kattula, B.T., "Comparative Study
of Prestressed Concrete Beams With and Without Bond", AQ.L
Journal, Vol. 68, No.2, February 1971, pp. 116-125.

1 5. Mojtahedi, S., and Gamble, W.L., "Ultimate Steel Stress in

Unbonded Prestressed Concrete", ASCE Journal of the Structural
Division, Vol. 104, No. ST7, July 1978, pp. 1159-1165.
 Prestressing 347

16. Pannell, F.N., "The Ultimate Moment of Resistance of Unbonded

Prestressed Concrete Beams", Magazine of Concrete Research, Vol.
21, No. 66, March 1969, pp. 43-54.

17. Tam, A., and Pannell, F.N., "The Ultimate Moment of Resistance of
Unbonded Partially Prestressed Reinforced Concrete Beams",
Magazine of Concrete Research, Vol. 28, No. 97, 1976, pp. 203-
208.

___ -...-18. Tam, A., "Ultimate Steel Stresses in Unbonded Prestressed

Concrete: Discussion", ASCE Journal of the Structural Division,
Vol. 105, No. ST9, September 1979, pp. 1862-1864.

~ 19. Naaman, A.E., and Siriaksorn, P., "Serviceability Based Design of

Partially Prestressed Beams", PC! Journal, Vol. 24, No. 2, March-
April 1979, pp. 64-89. ·;

20. Naaman, A.E., "Partial Prestressing in the Rehabilitation of

Concrete Bridges", in Proceedings of US-European Workshop on
Bridge Evaluation, Repair, and Rehabilitation, A.S. Nowak and E.
Absi, Editors, The University of Michigan, 1987, pp. 391-406.

\
 348 Naaman

APPENDIX I

The solution equations for the cracked section, assuming linear elastic behavior, are given below. For a given
moment M, the following equations lead to the stresses in the component materials.

2AsEs 2A'sE's 2ApsEps ~ {

+ [ 2 (b- bw) hf + ~ + --- + Oc - (Epe + Oc•ce)
-c Ec Ec M

(21)

Solve above equation for the value of c then compute:

(22)

fs ~
Ec fct (~)
c (23)

fps Eps (<pe + Oc•cel + Oc ~f

Ec ct (~)
c (24)

'ps
h
~ (25)

's
2
Eg (26)

f' s - E's
Ec fct (£..:.£)
c (27)

c' s !.i
E's (28)

NOTE: a) F = Apsfps: fps (thus F) varies with the applied moment M.

b) For rectangular sections b = bw: if no compressive reinforcement is used then A's) = o.
c) Sign convention; for concrete (compression +; tension -); for steel (tension +; compression -).
 Prestressing 349

TABLE 1
Expressions for the Strain Reduction
Coefficient n for the Uncracked State
---------------------------.-------------------------------
Loading Type and Strain Reduction Coefficient:
Tendon Profile Uncracked State

Uniform Load and Q 2

3
Straight Tendons
r--------------------------+------------------------------- -
Uniform Load and Q
5 ~
12 + 4 em
Single Draping Point
r---------------------------r---------------------------
Uniform Load and 0 _!!_ 2_~
15 + 15 em
Parabolic Tendon I
I
r----------------------r- --
Concentrated Midspan Load 1 n
2
1
and Straight Tendons '
-------------------------------r~--------~-------------
1
--
Concentrated Midspan Load n ~
3 + 6 em
~~-nd__S_ingle_D__r':_P_i_ng__Poi~---k ---
Concentrated Midspan Load ! n = 152 + 12
~
em
and Parabolic Tendon
r----------------------1--------------- --
2
Third Point Loads and n 1- (l
3
Straight Tendons
---------------------------r-------------------------
23 1 3
--
Third Point Loads and n ~
54 + 54 em
Single Draping Point
---------------------------r--------------------------
44 1 0
---
Third Point Loads and Q ~
81 + 81 em
Parabolic Tendons
~--------------------------L------------------------------- -
350 Naaman

TABLE 2
Solution Equations for the Uncracked State

Loading Solution Equations Equation

Case Number

fps • fpe 11

(Fe+ Mo) 12

13

n (M-Mo) eo
fps = fpe+ S 14
I Eps + Aps (r2 + e~)n

(F + M)
f.p1Aps [
1 +
eo ( d s - y 1) 1 M ( d s - y 1)
15
fes - r-c r2 ] - I
Mo<M<Mcr
~
fs = Ec fes 16

fet • ~[
Ac 1 - e...!LJ
kb + ~
I 17

Mer • (Mer)e + aMer 18

(F +Mer) (Mer)e • Apsfpe (eo+~) + 7.5ltJ ...Jf; 19

Apseo (eo+~) [ (Mer)e- Mo]

aMer • 20
~ I ~+ Aps (r2 - e 0 Zt)

NOTE: a) F - Apsfps; fps (thus F) varies with the applied moment M.

b) Sign convention: for concrete (compression +; tension -);

for steel (tension +; compression -).
 Prestressing 351

TABLE 3
Expressions for the Strain Reduction
Coefficient Oc for the Cracked State

Loading Type and Strain Reduction Coefficient:

Tendon Profile Cracked State

Uniform Load and nc = n..!il

lg + (1 - Y:) lf- 13 !2_)
J3
Straight Tendons

Uniform Load and nc = n..!il + (1 - k.t)[ !_ ~ 1.2 -

lg lg 2 em J2
Single Draping Point

.!_ ~ ~ !s:. 112.

OR
4 em J4 + I - 2 J2

Concentrated Midspan Load

and Parabolic Tendon -
1__!._2. 1 J4
J3 + 4 ~]
3

Uniform Load and nc = n~+

I
(1 - r:)[ em~(Ll2_ -
lg 3 J3

Parabolic Tendon L .!.2_) !s:. - 2 __!._2.

5 J5 + I 3 J3 +
1 !2_]
J5 5

Concentrated Midspan Load nc = n..kr + (1 - kr)

lg
[ l_c_I -
1 ~]
J2
lg 2
and Straight Tendons

Concentrated Midspan Load nc = n f: (1 -\';) [ l~) (1

+ - !_ 1£_]
2 I

and Single Draping Point + (.!.2 _ L3 I£_)

I
(~
em
_ 1) I~]
• Refer to Figures 2 and 4 for Notation.
352 Naaman

M
E
0
F
t Ultimate
Cracked

Elastic Cracked

A Elastic Uncracked
I
I

~
Figure 1
Schematic moment deflection curve of
an externally prestressed beam

LOADING TENDON PROFILE

l I
w
* **** * ***
1

t t
I
t t
I
~p

8 -l:F-~-4
5
2
I
t t
I t t

~a1.ff31=}
3
1t I N-RABOLIC ---1
I t t t

Figure 2
Typical loadings and tendon profiles
considered in the computation of the strain
reduction coefficients n and nc
 Prestressing 353

~ Compression

e>v; 'j

reference state
Fe + MD

I I
:.. ..:I bonded
~€ :
ps :
I

I.; ..........: unbonded

I
I
I

Figure 3
Strain differential between bonded and
Unbonded tendons in bending - Uncracked state

~ I

Figure 4
Typical representation of elastic uncracked,
elastic cracked, and idealized elastic cracked beam
354 Naaman

0.50.---------------------.,
UNIFORM LOADING- SINGLE DRAPING POINT PROFILE

c
w
~
~ 0.30
a:
0
c:(
CJ 0.20
w
lett lg. 0.2
-11:1-
0== ......
lcr/lg = 0.4
0.10 -a- lcti lg .. 0.6

...
+ let/ lg. 0.8
leti lg • 1.0
0.00 +--..----,---r--r---r---r---.,.---r--.---1
0.0 0.2 0.4 0.6 0.8 1.0

lc/L

Figure 5
Typical illustration of the variation of nc

~ Compression

- --------~-r--- F + M
(M >Mer)
dps
,___- reference state
e Fe+ MD

.._I_ _..... :I
ce £
pe
H
I I

:.....r---...i bonded
I
I
A £ I
Ll. ps:
I
~unbonded
I
I
I

Figure 6
Strain diagrams from the uncracked to
the cracked state
 SP 120-17

Ultimate Behavior of Precast Segmental

Box Girders with External Tendons
by J. Muller and Y. Gauthier

synopsis The concept of precast segmental

construction with external tendons has been developed
extensively si;nce 1978, starting with the
construction of the Long Key Bridge (Florida,
U.S.A.). Since this first experience, many other
structures (more than 5,500,000 SF of deck) have been
designed and successfully built using the same
method. /
The performance of all bridges now in operation has
been excellent. However, some questions were raised
1n the minds of Engineers, unfamiliar with the
method, as to the behaviour of structures prestressed
with external tendons beyond the range of design
loads (serviceability limit state). Because
continuous reinforcement is not usually provided
across the match cast joints between segments,
concern was expressed that adequate ultimate
behaviour and sufficient strength could not be
obtained.
In order to provide a satisfactory answer to these
legitimate questions, a special computer program
(DEFLECT) has been developed to analyse accurately
the response of the structures prestressed by
external tendons. Moreover, several tests are
available to confirm the behaviour of such struc1:ures
while verifying the validity of the DEFLECT computer
program. This design tool has been used to prE~dict
the structural behaviour of simply supported and
continuous structures beyond joint opening, up to
ultimate capacity of the girders with and without
thermal loads. Several different prestressing methods
have been analyzed with different bonding conditions
at the point of deviation of the external tendons. It
was found systematically that stru~t'llres_ prestressed
with either inb:~rria1 or external tendons.- t:rehave
essentially in the same way at . .all. loading ...stage.s up
to ultimate.
Ite\!Woi-dS:· bending moments; box beams; bridges (structures) ;
c:orrputer programs; moment-curvature relationship; I1Q§t::
tensioning; precast concrete; prestressing steels; segmental
construction; serviceability; stress-strain relationships;
structural design; unbonded prestressing

355
356 Muller and Gauthier

Honorary ACI member, Jean M. Muller is presently

Technical Director of Jean Muller International Inc.
of Paris, France and of Jean Muller International
Inc. of San Diego, California, USA. He has been
involved actively in the design of innovative
concrete bridges for the past 40 years.

ACI member, Yves Gauthier is presently Assistant

Technical Director of Jean Muller International Inc.
of San Diego, California, USA. He was actively
involved in the structural analysis of concrete
bridges with external prestressing which forms the
basis of this article.

l.Scope
The concept of precast segmental construction with
external tendons has been developed extensively since
1978, starting with the construction of the Long Key
bridge.
Since the experience of the four segmental bridges in
the Florida Keys, representing a deck area of
2,200,000 SF, many other structures have been
designed and successfully built using the same
method.
The following list of structures also includes
projects still under construction :
 Prestressing 357

Wiscasset, ME 122,000 SF
Dauphin Island, AL 165,000 SF
Sunshine Skyway Bridge, FL 735,000 SF
Escatawpa River,MS 240,000 SF
Albemarle Sound, NC 620,000 SF
I-110 in Biloxi, MS 615,000 SF
I-170 in Glenwood Canyon,CO 75,000 SF
Wando River, sc 740,000 SF
James River, VA 540,000 SF
Neches River, TX 528,000 SF
San Antonio Y, TX 1,300,000 SF

The performance of all bridges now in operation has

been excellent. However, some questions were raised
in the minds of Engineers, unfamiliar with the
method, as to the behaviour of structures prestressed
with external tendons beyond the range of design
loads (serviceability limit state). Because
continuous reinforcement is not usually provided
across the match cast joints between segments,
concern was expressed that adequate ultimate
behaviour and sufficient strength could not be
obtained.
This report provides a satisfactory .answer to these
legitimate questions.
Several tests are available to confirm the behaviour
of girders prestressed with external tendons. They
were used to verify the validity of a computer
program developed to analyze accurately the behaviour
of these structures.
It was found systematically that structures
prestressed with either internal or external tendons
behave essentially in the same way at all loading
stages up to ultimate.
A design tool now exists to predict accurately the
ultimate strength of a structure with external
prestressing.
358 Muller and Gauthier

2. Presentation of the DEFLECT computer program

The "DEFLECT" computer program allows analysis of
structures beyond the decompression of the match cast
joints. The computation is based upon the
relationship between bending moment and beam
curvature.
2.1. Methodology
In the precast segmental bridges, when the loads
increase beyond the service loads, usually there is
no crack occurring between joints of the segment, and
all the cracks are concentrated at the joint
location, thus inducing opening of the joints (Fig
1) •

FIG 1.
Crack pattern in segmental bridge

This fact is especially t:rue for precast segmental

bridges with dry joints (no expoxy glue in the
joints).
When the joint opens, the beam rigidity is modified.
The rigidity of the structure depends on the
deformability of the precast segment between two
adjacent joints.
To predict the beam rigisiity, it is necessary to
know the relationship between bending moment and
curvature or bending moment and rotation at joints.
Consequently, the problem will be solved if it is
possible to calculate the rotation dw occurring at a
joint for a given pair of actions, N (axial force)
 Prestressing 359

and M (bending moment), which induces a specified

opening of the segment joint as shown on Fig 2.

dw

FIG 2
NOTATION AT JOINTS
The deformability of a typical segment can be
obtained by means of a finite element analysis in
order to account for the shape of the section, the
length of the precast segments and the height of the
section in compression at the joints.
The segment is divided into volume elements as shown
in Fig 3.:

FIG 3
FINITE ELEMENT MODEL

All the calculations are carried out assuming that

the stress distribution at joint is linear (contact
360 Muller and Gauthier

zone is plane) and the concrete in tension between

two opened joints is not cracked.
step 1

For a given neutral axis, a unit translation and then

a unit rotation are applied to all the nodes of the
joints which are above the neutral axis. The results
of both calculations are combined in such a way that
the resulting displacements of all the nodes above
the neutral axis are a rotation dw around the neutral
axis.

Nl
I Fig 4
·-·-r·cr••••d
.mal Aala
. C.O.G.
s ... ,•• Principle of computation

step 2

The computer program can calculate for all nodes Pi

the axial force Ni resulting from the applied
displacements. Assuming Zg to be the altitude of the
center of gravity, one can write :
N = L: Ni
Mg =4 Ni (Zi - Zg)
The relationship between the rotation dw and the
applied load characterized by the pair of valwas, N
(axial) and M (moment), can be obtained by changing
the depth of the neutral axis.

For convenience, this is expressed between the two

parameters M/N and Edw/N (E Modulm; of
elasticity).

In figure 5 , the moment-curvature relationship is

given for a typical segment of the Sunshine Skyway
bridge approaches.

If the resultant force stays within the central core,

the response is linear and the slope of the diagram
is equal to I (I= bending inertia).
If the eccentricity increases, the curvature
increases much faster but is always perfectly
 Prestressing 361

..
~
u
II

.I .
;.
~ N
,.;
i.
u
~

SUNSHINE SKYWAY BRIDGE APPROACHES

Fig: 5 MOMENT /CURVATURE RELATIONSHIP
FOR TYPICAL SEGMENT
362 Muller and Gauthier

reversible up to a point very close to ultimate

state.
At ultimate state, depending on the sign of the
bending moment, the curvature is 450 or 80 times
greater than for the section totally under
compression.
It must be pointed out that the analysis is conducted
assuming that the materials remain in the elastic
range.
starting with the bending moment-curvature
relationship, the "DEFLECT" program determines the
equilibrium of the segments along with the post-
tensioning steel overstress under the loads applied
to the structure. The prestressing steel overstress
depends upon the geometry of the tendons and the
bonding conditions prevailing at the deviation
blocks.
The stress strain diagram for post-tensioning steel
is shown in fig 6.
0/o GUTS ( 0/o Guarantood Ultimate Tenelle Shength)

£1 :< 4 0715 000 KSF

f 1 a 38 880 KSF ( 270 KSI)

FIG 6
Example of P/T steel stress-strain diagram

The stress distribution in the structure is

calculated by iteration until satisfactory accuracy
is obtained. The methodology also allows the analysis
of loads combined with temperature gradients (linear
or non-linear).
The validity of the DEFLECT program has been checked
by comparing it with available test data for similar
structures.
2.2. Comparison with available test data
2.2.1. CEBTP Test
A test was conducted at Centre Experimental de
Batiments et Travaux Publics (C.E.B.T.P.) in France
 Prestressing 363

6.000m (19'-8 14"1

0.100.1

TESTING OF SEGMENTAL GIRDERS

WITH EXTERNAL TENDONS

..
0
0
..J

...
0
00

.
..J
Q.
Q. 50

0
DEFLECTION OF EXTERNAL TENDON GIRDER c.E.I.T.P. TEST

THEORETICAL DEFLECTION FROM DEFLECT

THEORETICAL OEFLECT10N Of INTERNAL TENDON GIRDER
FROM DEFLECT

0 ~------~------~--------~------~--------~----~--~
0 0.25 0.50 0.7!1 1.00 1.25
1.4T 1.14
DEFLECTION ( ln.J

C.E.B.T.P. TEST
FIG 7
364 Muller and Gauthier

on a precast segmental beam with dry joints and

external post-tensioning tendons.
Figure 7 shows the test set-up and results. The
plotted deformations obtained from the test have been
compared to the theoretical "DEFLECT" calculations .
Both of the curves are consistent except at failure
since the computer program uses the elasticity theory
and cannot account for the concrete plasticity at the
ultimate stage. However, the "DEFLECT" analysis
underestimates the ultimate capacity of the beam.
It should be mentioned that at ultimate loading there
was a 46 ksi overstress in the prestressing steel as
compared with the 15 ksi overstress allowed by the
existing AASHTO code.

Another computer run was made for a beam with exactly

the same geometry and prestress tendons as the test
beam with external tendons, but it was assumed that
the tendons were fully bonded on the whole length.
The differences between the three curves are
insignificant at all loading stages up to and
including ultimate capacity.

2.2.2.CTL(Construction Technology Laboratories) Test

To ascertain the suitability of the concepts of
precast segmental construction with external tendons
to sustain seismic loads, a comprehensive testing
program was implemented with particular attention
focused on redundancy and ductility of the structure.
For purposes of comparison, three identical beams
were tested and instrumented first up to design load
and in the elastic range and then loaded to ultimate
capacity.
This test has already been described in previous
articles. (P.C.I. Journal March/April 1987 page 86 :
"Testing of segmental concrete girders with External

r
Tendons" by K. sowlat and B.G. Rabbat).
The following conclusions may be drawn from the test
results :
- Beyond the design stage, the unbonded beam is more
flexible than the bonded one.
- A complete elastic recovery was experienced after
loading the beams up to 1.7 times the design load in
spite of the substantial joint opening of the
critical joints (0.14 in).

The unbonded beams develop an ultimate moment,

which was the same as that of the bonded beam.
 Prestressing 365

Q Applied Lood (Kips)

_.. ......
__,..-·
. -~
__.~-~
/
;·
J
.!J1 -•-ou~;"'OON Slf un-. ~
··-•··THfORETil"...&l 09'1 "-liON '""" D<Fl.[CT
9:!111" ·!liT 1 I•••U

Q Applied
. ood (Kips) ' " 1
Deflecticfr1 (Inches)

~----·
.
k~~
_.-

/.·'
_£:
'

r
lj --Ol:FUCTION OF' UT!NW..
.. _._.f... ,_Tto ""' ~~
~N

~
GR21:1t • TEJT

"'' ·~
I l!!!l!l:lltl

. Oef I ection ( Inches)

...

~
[!I s.e--• ,._....
e~= s'::.~h:.
T....... I. 2 K!
J """' •.••

C.T.L TEST
FIG 8
366 Muller and Gauthier

These experimental studies are confirmed by computer

analysis.
Figure 8 shows the test set up for the unbonded
beams and the comparison between measured beam
deflection and theoretical deflection, calculated by
means of the "DEFLECT" program for two tests
corresponding to two different post-tensioning
patterns.
Again, the plotted curves are quite consistent,
confirming the "DEFLECT" computer program to be a
satisfactory tool to predict the ultimate capacity of
precast segmental structures.
3.Analysis of Precast Segmental Box Girders at
Service Load Design
The superstructure of the Sunshine Skyway Bridge high
level approaches has been selected as representing an
excellent current example of box girder bridges with
moderate spans. A typical unit consists of 8 spans of
135 ft.
Figure 9 gives the dimensions and the properties of
the typical cross sections and the prestressing
layout for the first and second spans. The other
spans are identical to Span 2, except that the 1 3/8"
P/T bars do not exist in the intermediate spans (Nos.
3 thru 6).
The bridge has been designed for a 18 •F linear
temperature gradient which induces, in continuous
bridges, similar tensile stresses in the bottom
flange as the complex gradients and relatively high
temperature differentials outlined in appendix A of
National Cooperative Highway Report 276.
The joint decompression has been checked for the
following load combinations (no tension allowed)

DL + (LL +I) (Dead Load + Live Load + Impact)

DL + Gradient (18 •F)
DL + (LL + I) + Gradient (9 •F)
The superstructure was also checked for two
additional combinations :
DL + Gradient (18 •F) + 1/2 (LL + I)
DL + Gradient (-9 •F) + (LL + I)
4. Analysis of Precast Segmental Box Girders with
External Tendons at Load Factor Design
The superstructure of the Sunshine Skyway approaches
has also been analyzed at load factor design by means
of the "DEFLECT" computer program.
 Prestressing 367

AREA ~ 56.25.,
/8 = 48J.5H
c ~ 2.478"
CP ~ 5.522"
0 • 73.512"
_,_ = 6.58"
0
TYPICAL CROSS SECTION

SPAN I
P.T. LAYOUT

SPAN 2
P.T. LAYOUT

SUNSHINE SKYWAY BRIDGE APPROACHES

FIG 9
368 Muller and Gauthier

The first span which is the most critical under

ultimate loads has been analyzed for the following
load combinations.
4.1. 1.3 DL + 2.17 (LL + Il See Figure 10

Under ultimate loads, two joints are open at mid-

span. The materials remain in the elastic range
with
fs 188 ksi (70% G.U.T.S.)
fc 319 ksf (40% f'c )

Two other joints are slightly open on either side of

the pier segment.
4.2. 1.3 DL + 0.5 GRADI + 2.17 (LL +Il See Figure IO

At ultimate load, we have also analyzed the effect of

a temperature gradient.
The comparison of the forces and of the deflections
with 4 .1.1 shows that the effect of the temperature
gradient does not modify significantly the behaviour
of the superstructure under ultimate loads.
fs 190 ksi (70% G.U.T.S.)
fc 355 ksf (45% f'c)

The maximun overstress in the external tendons is

21 ksi.

4.3. 1.3 DL + 3.49 (LL +Il See Figure 10

This load combination corresponds to the ultimate

capacity of Span 1. The plot of the structure shows
that the ultimate capacity of Span 1 is obtained when
joints at mid-span and at the pier are simultaneously
fully opened.
The stress in the concrete is in the neighbourhood of
85% f'c but the P/T steel stress stays smaller than
217 ksi (80% G.U.T.S.) with a maximum 50 ksi
overstress.
Based upon the comparison with available test data,
"DEFLECT" underestimates slightly the ultimate
capacity of the beam, since the calculations assume
an elastic response.
4.4. 1.3 DL + 2.17 (LL + Il, with no Bond in
Deviation Blocks See Figure 11
The P/T tendons are encased in steelpipe ducts at the
deviation block location. Usually the bond in the
duct allows the transfer of overstress. In order to
 Prestressing 369

SPAN I-JOINT 6
MAX. ULTIMA T£ LOAD /pill' ASSHTOI
1.3 OL • 2./7 (LL •II

13!5' 135'
SPAN I - JOINT 6
MAX. ULTIMA T£ LOAD /pill' ASSHTOJ
1.3DL •0.5GRAD•2.17(LL.[J

-...
!!!
~ 13!5' 13!5'
'i' 'i' :::u SPAN 1- JOINT 6
MAX. ULTIMATE: CAPACITY
1.3 DL+3.49 (LL +II

SUNSHINE SKYWAY BRIDGE APPROACHES

ULTIMATE BEHAVIOR
FIG 10
370 Muller and Gauthier

show the safety factor of the superstructure, we have

also analyzed the superstructure, assuming that there
was no bond in deviation saddles.
Figure 11 shows the state of the first span under
ultimate loads. Obviously the structure is more
flexible but the stresses are still within the
elastic range.
fs 172 ksi{GG% G.U.T.S.) (overstress 11. 6Ksi)
fc 454 ksf {57% f'c)

SPAN /-JOINT 6
MAX. Ul.. TIMA T£ LOAD O>er ASSHTOI
!3 DL • 2.17 ( LL f/)
NO BOND IN DEVIATION 8...0CKS
He • height in comprflsston

Fig: II SI:.INSHINE SKYWAY BRIPGE APPROACHES

ULTIMATE BEHAVIOR
 Prestressing 371

5. Structural Behaviour of Precast Segmental Bridges

with External Tendons beyond the Joint Opening

The aforementioned load combinations have

demonstrated the satisfactory flsxural behaviour of
precast segmental bridges at load factor design (per
AASHTO) and at ultimate capacity.

It is important to understand the mechanism of load

transfer after opening of the joints.

Figure 12 shows the trajectory of the pressure lines

corresponding to different load combinations for Span
1. In fact it is obvious that an arch effect is
developed in the structures, thus allowing the shear
forces to be directly transferred to the bearings.

PRESSURE LIN£

7.17'../J ..

~ ~ 8. ~ .
~ l! 8;g 8
.,~ .iii"' ~
i
!:! !!
i l i i :l 1" 1
SPAN 1 PRESSURE UNE AT ULTIMATE LOAD !PER AASHTO!
PRESSURE LIN£

Il ~
~
~
l
~
i
i1
SPAN 1 PRESSURE UNE.AT ULTIMATE CAPACITY

Fig: 12 PRESSURE LINES

372 Muller and Gauthier

6. Comparison of Different Prestressing Methods for a

100 ft Simple Beam
Another interesting experience reported in this
paragraph was to consider a 100 ft simple beam with
the same cross section and prestress as the outside
span of the Skyway approaches. Cross section and
properties are shown in Figure 9. Prestress consists
of 2 x 76 = 152 STRANDS 0.5 11 diameter giving a steel
area of 23.2 si = 0.161 sf.
The same beam was successively assumed to have the
following characteristics :
- PO : External prestress with 4 deviation saddles
and 10 1 long segments.
- Pl : External prestress with deviation saddles
at 10 1 intervals
- P2 : External prestress with deviation saddles
at 5 1 intervals
- P3 : Internal prestress modelized with deviation
saddles at 1 1 intervals.
The results are very interesting and somewhat
surprising (see Fig 13) . The so-called bonded beam
(P3), so often taken as a reference of maximum
performance, does not show the largest deflection
near ultimate. A beam with a finite number of
deviation andjor anchor blocks allows more
flexibility to develop at ultimate.
As a conclusion, all beams behave much alike.
PI P3 P2 PO
3"-2!5
35
34
33
3Z
I I I

r::s
" I I I
30
jllllilfH±t
~ I
~

-
100' Soon s.,.. I
0
100'
-"'
.."'
0
0
-'
25 !'Oj P£YIAT!JN ILOCK 1PA. AS P!:DEVIATION a.oc!( SPa. AT 5'

20
I
19 JOINTS OPEN1NG

Pt: DEVIATION ltLOCK SPA. AT to'

100' Span i
P3: DEVIATION Ill.()()( Sflll. AT I'
, I
15 0~------0~.5-------,o------~-+~~~--~-~~--­

OEFLECTIC'N (Ft.)
...
Fig 13
Comparison of several prestressing methods
 Prestressing 373

7. Second order deflection

Speci~~ care must be taken in the analysis of the
second order deflections since, when the beam is
close to failure, a slight increase of the
eccentricity · magnifies the beam deflection. T~e
increased deflection induces a loss of post
tensioning tendon efficiency if the tendon does not
follow the concrete deflection.

Loll of Eccentricity

FIG 14 shows the superiority of a beam with

properly distributed deviation blocks.

8. Conclusion
Prestressing with external tendons has proved to be a
very practical and efficient tool in the field of
concrete bridge particularly with precast segmental
construction. It is not i:t11::~!1.Qed ±o....r.epl.ace internal
pre_f:!ltressing in all st,rl,lct_l,!res, bu!- rather complement
the more conventional prestressing methods with
bo1'lded · tendons .
In term of structural behaviour and design methods,
the experience gained during a period of now more
than ten years has proven the method of external
prestressing to be essentially equivalent to the
method of internal prestressing. Design tools are now
available to analyze and predict the behaviour of
structures pretressed with external tendons at all
loading stages from serviceability limit state up to
ultimate limit state and even actual ultimate
capacity.
It is anticipated that external prestressing will be
used more extensively in the future in concrete
construction, but also in steel or composite concrete
and steel structures.
 SP 120-18

Externally Prestressed Bridges

by J. Eibl

Synopsis: Experience gained in the design of several

externally prestressed bridges is reported. New
cables especially developed by contractors are dis-
cussed. Also the main experimental results gained at
the author's institute for reinforced concrete are
g1ven.

Keywords: bridges (structures); prestressed concrete;

prestressing; prestressing steels; structural design; tests;
unhanded prestressing

375
376 Eibl

Dr. Engineer Josef Eibl is the Professor of Struc-

tures, Department of Structural Engineering, Insti-
tute for Structures and Building Materials Technol-
ogy, University of Karlsruhe, West Germany.

1. INTRODUCTION

Experience has shown that the grouting of prestress-

ing cables needs great care if later damages are to
be avoided. As the latter cannot safely be guaran-
teed in every case the question has been raised in
Germany whether the inspection of cables and their
replacement, if necessary, could be provided in
future bridge design. Such an intention leads to
cables without bond. In principle they could be
arranged within the webs of T-beams or hollow box
cross sections as in the case of bonded cables.
Striving however for more robust bridges it seems
reasonable to place the cables outside the webs, so
that an easier and safer concreting of this struc-
tural member can be achieved. As we know, bundles of
prestressing cables within the webs have led to con-
siderable damage in the past.
Because the strain increase in the ultimate limit
state is smaller with unbonded cables than with
grouted ones, one usually needs more mild reinforce-
ment within the cross section. This however can also
help to gain more durable structures. In the past it
has been demonstrated that even in cases of high lev-
els of prestressing a reasonable amount of mild rein-
forcement has to be used to achieve controlled crack-
ing.
This type of reasoning, which has already led to a
remarkable number of external prestressed bridges in
France, is among others greatly supported by BRUGGEL-
ING [4,6,7,8] (Fig. la), who more or less demands a
reinforced bridge with an additional favourable
stress state generated by prestressing.
Starting from the same basis of consideration the
author has designed, in close cooperation with the
German Autobahn Authorities and the German Railways,
five bridges which are under construction now. In
the case of a Swiss bridge, which has
 Prestressing 377

already been completed, the author acted as a scientific con--

sultant to the Swiss Authorities.

2. DESIGN ASPECTS

Starting from this basis as a first alternative one could ar-

range the cables as shown in Fig. lb, which allows easy
placement, easy replacement if necessary and also the possibi-
lity of using the same cables at different places and for dif-
ferent purposes during the construction process without high
labour costs.

11111111111 tttttttttt 11111111111

II II I I I ---IIIII-- I Ill I I I

Fig. 1
378 Eibl

A structurally more effective arrangement of cables is shown

in Fig. lc where vertical forces are applied to the concrete
section at the bends of the cables.

In both cases a number of anchoring devices has to be placed

within the cross section. These are necessary to change the
direction of cables in horizontal or vertical direction and for
permanently or temporarily anchoring in case of special erec-
tion procedures such as incremental launching e.g. The transfer
of such sometimes great anchorage forces to the cross section
has been carefully analyzed by means of threedimensional F.E.

Section 8-8
@ @

Section A-A

8 8 8 8
.1._ __ 1 .t. _ __ j

II
CD

Fig. 2
 Prestressing 379

investigations in order to avoid great tensile stresses, which

might occur in the webs.

Fig. 2 d demonstrates bending which results from anchoring in

a vertical beam which was only fixed at the neighbouring webs
as shown in Fig. 2 c. In Fig. 2 e the anchoring beam was only
fixed at the top and bottom slab, while Fig. 2 f gives the
best solution with regard to minimal bending stresses. In this
case the vertical beam with a rather big dimension in the
bridge's longitudinal direction is fixed at top and bottom and
has an additional rather thin shear connection to the
neigbouring web.

The increase of cable forces due to applied live loads or to

increased ultimate loads as well as the relative displace-
ments at the saddle are small (Fig. 3). According to:

L L

J[cc(s)- c.,] ds = 0
0
€11 =~ J
0
cc(s) ds
( 1)

au A = '7•[e,( S) - e,j ds
11=0

the strain difference &c (s)-& s cancels nearly out

along the path of integration.

Of great importance is the type of prestressing cables, which

are to be used. In the case of external prestressing a homoge-
neous distribution of prestressing steel over the concrete area
is not relevant and therefore one will prefer rather heavy
cable units. The upper limit is dictated by the handling
process at erection. A force of 2-3 MN seems to be reasonable.
380 Eibl

Bridge

Moment due to live load

Fig . 3

In Germany three cable types have been especially developed

for this purpose (Fig. 4). Others will be tested in the future .

PE-Tube PE - Tube PE - Tube

Prestressing Prestressing
Wires Wires

Type 1 Type 2 Type 3

Fig. 4
 Prestressing 381

Type 1 and 2 of BBRV rsp. SUSPA cables consist of 7 nm dia.

2
wires with fy/fult = 1470/1670 N/nm and are prestressed
to 0. 7 f ult" This is different from bonded cables which can
only be stressed up to 0.55 f according to German stan-
u1t
dards. The reason is that unbonded cables with their small
stress variation under traffic loads may be treated as
statically stressed.

The wires are arranged in a plastic tube surrounded by grease

or mortar. Type 2 contains an additional metal cover inside
the plastic tube which does not hinder cable transport on
coils. This is necessary to avoid heavy deformation of the
plastic tube at saddle points as our experiments have shown.
The anchorage system is the BBRV rsp. SUSPA system. When
cables are filled by grease the complete cable is manufactured
and sealed in the plant thus guaranteeing high quality at the
site.

Type 3 has been developed by Dyckerhoff & Widmann and consists

of the usual monostrands surrounded by mortar in a plastic
tube. At saddle points specially designed holders guarantee
the correct position of the strands thus avoiding high trans-
verse pressure that otherwise may damage the surrounding
plastic tube. An effective anchorage system allows to replace
every single monostrand.

3. EXPERIMENTAL RESULTS

All cable types have been tested experimentally to check

6
whether they can endure 2.5*10 load cycles under a live
2
load amplitude of 35 N/mm and a relative displacement at
the saddle of 0.6 nm (see Fig. 5). The cable was supported at
the saddle on torus-shaped metal half-shells in the case of
Type 1 and 2 tendons. Type 3 was only supported by a special
mortar bed moulded by formwork on top of the concrete saddle.
382 Eibl

The radius of curvature at the saddle was 4 m. The change of

inclination varied from 1 to 6 degrees. In the case of the
BBRV cable used for the bridge in Switzerland temperature
variations between -l5°C and +35°C have also been applied
during tests.

? ~ /

Fig. 5

All relative displacement between the different elements of

the cross section have been recorded.

The main results gained by these tests may be summarized as

follows:

-All cables showed satisfactory behaviour

- At angles > 1 degree the saddle acted as a fix -
point for the cable
- Friction was about 1% at first prestressing
- The grease filled Type 1 cables showed a noticeable
deformation on the inner surface of the plastic
tube due to the transverse pressure

4. APPLICATIONS

In Fig. 6, a sketch of the Bridge at Preonzo Claro which has a

maximum span of 62.5 m, is given. It was originally designed by
the consulting office of Guzzi in ZUrich. In this case BBRV ca-
bles with a capacity of 2.35 HN, 3.7 HN, and 4.6 HN were used .
 Prestressing 383

The bridge has nearly been completed. Experience showed that

4.6 MN cables should be an upper limit with regard to handling
i nside the hollow box cross section.

11 1 1'111 :v-
I I
7 X 44,50 35.50
509,50m

Bridge of Preonzo Claro (Autostrada Chiasso -San Gottardo)

Fi g. 6

Longitud inal Section

Cable -Layout

Fig. 7

In Fig . 7 the Wintroper Talbrilcke, an Autobahn bridge with two

parallel decks is shown . It will be built by the incremental
launching method as developed by LEONHARDT . In this sketch the
cable layout is also drawn.
384 Eibl

A similar Autobahn bridge - Berbke -with maximum spans of 45 m

and a length of 298.50 m is also under construct i on and is
being built using the incremental launching system. This bridge
is curved in its ground plan with a radius of 2350 m.

Fig. 8 shows two different systems which will be applied to

three single-span bridges for the German Railways. Their erec -
tion will start by the end of this year.

1. 44.aam .1

I. 44.00m .I
Fig. 8

In the case of the Autobahn bridges, the necessary prestress ing

force was determined by the demand that "c at the bounda-
ries should be zero under dead load. One third of live load ,
plus the usual forces associated with ± 5°C between the cables
and the neighbouring concrete had to be considered, a deviaton
from the usual regulations for bridges with bonded cab l es .
Crack width was to be checked by means of a given specificat ion
demanding a calculated crack width of 0.15 mm under increased
live loads. This was to control service behaviour.

Also the bridges must be able to support half of the traffic

load associated with the full service conditions without two
cables so that easy replacement of cables can take place with
only small traffic restrictions.
 Prestressing 385

At the ultimate state an increase of the prestressing force was

allowed only if detailed analysis of the cracked state was
done .

In the case of the railway bridges, for the first time, after a
thorough investigation decompression already under half of the
live load was allowed. Until now full prestress was always been
demanded by the German Railways.

Fig . 9

Fig . 9 finally shows a new design of the author for a composite

externally prestressed railway bridge, which is still in the
biddingprocess and consists of a top and bottom concrete slab
and steel truss webs. It is in so far an extension of external
386 Eibl

prestressing as the cable sag is now increased by anchoring

the cables outside the cross section at vertical members of
the steel structure. This allows the webs to be resolved to
steel truss girders which may also serve as scaffolding
girders during the erection process when the upper and lower
concrete deck is poured.

5. ADVANTAGES AND DISADVANTAGES

Regarding the little experience the author could gather, the

following advantages and disadvantages are seen.
If the cables cannot be applied external to the cross section,
usually only a lower lever arm for prestressing is available
than in the case of bonded prestressing and with additional
loads in the ultimate state more mild reinforcement is neces-
sary. This however led to only half of the usual calculated
crack width. Intensive comparison calculations show that with
less prestressing force and more mild reinforcement, the costs
for the entire reinforcement calculated on the basis of a cost
relation of prestressed steel/mild steel = 4.5 have an increase
of 10%.
An advantage is that cable manufacturing especially with grease
filled cables manufactured off the site quarantee greater
quality. External cables allowed better concreting of the webs
and thus thinner webs. In addition new types of steel webs may
be used. The possibility of easy inspection and replacement
will certainly prevent some of the past bad experriences. With
a higher amount of mild steel partial prestressing will lead
to better crack control. Restrictions in present regulations
that do not allow cracks to cross prestressing cables are no
longer necessary. De-icing salt is practically of no influence
to the cables. The cost increase of about 3% for the whole
bridge is negligible. These developments may also lead to new
structural types, as is shown with the last example.
 Prestressing 387

REFERENCES

[1] voss, w. Vorspannung ohn~ Verbund am Beispiel

der TalbrUcke "Berbke" im Zuge der
A 46, Arbeitstagung BrUcken und In-
genieurbau, Tagungsbericht Regens-
burg 1986

[2] VIRLOGEUX, M. External Prestressing of Concrete,

FIP-KongreB New Dehli, FIP Notes
19B7/2

[3] VIRLOGEUX, M. Die externe Betonvorspannung, Be-

ton- und Stahlbetonbau (83) 1988

[4] BRUGGELING, A.S.G. Structural Concrete Science into

Practice, HERON, Vol. 32 No. 2., 1987

[5] WITTFOTH, H. Wohin entwickelt sich der Spannbeton-

bau? Beton- und Stahlbetonbau 81
(1986)

[6] BRUGGELING, A.S.G. Uitwendige voorspanning, CONSTRUK-

TIEF BETON XIII

[7] BRUGGF.LING, A.S.G. KONSTRUKTIEF BETON: WETTENSCHAP EN

PRAKTIJK, Kathol ieke Universitteit
Leuven, Fakul tei t Toegepaste Wetten-
schappen, leuven 1985

[8] BRUGGELING, A.S.G. External Cables, State of the Art

Report, Conf. Partial Prestressed
Concrete Structures, Bratislava 1988
 SP 120-19

Finite Element Analysis of

Externally Post-Tensioned
Segmental Box Girder Construction
by M.E. Kreger, G.L. Fenves, and K.C. El-Habr

Synopsis: A nonlinear finite element analysis was conducted to examine the full range
of behavior of segmentally precast box-girder construction with external post-tensioning
tendons. A primary objective of the study was to examine the effect of dry joints (without
epoxy) on the strength and ductility of box girder construction. A secondary consideration
was the influence of supplemental bonded internal tendons on the behavior of t he structure.

Keywords: box beams; bridges (structures); finite element

rrethod; post-tensioni.rg; precast concrete; prestressinJ;
prestressinJ steels; seqrrental construction; structural analysis;
unborrled prestressinJ

389
390 Kreger, Fenves, and El-Habr

ACI member Michael E. Kreger is an Assistant Professor of Civil Engineering at The

University of Texas at Austin. He received BS, MS and Ph.D. (1983) degrees from the
University of Illinois at Urbana-Champaign. He is a member of ACI Committee 318-C,
Analysis, Serviceability and Safety, ACI-ASCE Committee 352, Joints and Connections,
and ACI Committee 215, Fatigue of Concrete. He has conducted research in the areas of
repair and strengthening of concrete structures, behavior of reinforced concrete buildings
subjected to seismic loads, and behavior of prestressed concrete bridge structures.
Gregory L. Fenves is an assistant professor of civil engineering at the University of
California at Berkeley. He obtained his Ph.D. in 1984 from the University of California at
Berkeley. His research interests include nonlinear finite element analysis, particularly for
earthquake response of concrete dams.
ACI member Kamal C. El-Habr is a structural designer for Whitman, Requardt and
Associates, Baltimore, Maryland. He received his BS and MS degrees from The University
of Texas at Austin in 1984 and 1988.

INTRODUCTION

Since the completion of segmental bridges in the 1950's and 1960's in Europe, and
in the 1970's in the United States, segmental prestressed concrete box-girder bridges have
become a predominant form of medium and long- span bridge construction (100 to 650
foot spans) that has been characterized by rapid evolution. Economic factors have contin-
ually suggested refinements in construction practices to increase productivity and minimize
construction time. Each of these developments in construction methods has required in-
novative design procedures. Principal among these are the relocation of tendon anchorage
zones, the use of multiple shear keys, the omission of epoxy from the joints, and the use of
tendons external to the webs and flanges.
A significant number of segmental prestressed concrete box girder bridges with ex-
ternal tendons and dry joints have already been constructed. Substantial economic and
construction time savings have been indicated for this type of construction. However,
relatively little analytical investigation has been undertaken to evaluate the behavior of
segmental bridges, incorporating the new developments, for all ranges of loads. The ana-
lytical evaluation of their behavior is desirable before additional construction proceeds.

FINITE ELEMENT FORMULATION

A segmental girder with external tendons is modeled with three types of finite el-
ements: a one-dimensional beam element for the segments, a tendon element connected
to the segments by rigid diaphragms, and a joint element that transmits forces between
two segments. Figure 1 illustrates a simple-span girder of two segments modeled with four
beam elements, three tendon elements, and one joint element.
 Prestressing 391

Beam Element

Segments are modeled by six degree-of-freedom beam elements. Transverse loading

in a vertical line of symmetry is assumed. The nonlinear behavior of the concrete and
passive reinforcement is approximately modeled by a series of fibers in the cross section
[1,2,3]. A uniaxial constitutive relationship is specified for each fiber (such as in Ref. 4
for concrete and a bilinear relationship for steel). Although a fiber model requires more
computation than a model of the beam section, moment- curvature-axial force relationships
do not have to be defined [3).
The strain in the element is

(1)

where Ll£ is the increment of strain from the current strain state, £ 0 , and all strains (and
stresses) are positive in compression. Neglecting shear deformations, the assumption of
plane sections gives

(2)

where Ll£a and Ll¢ are increments of axial strain and curvature, respectively. In the finite
element approximations these increments are expressed as

{~;} = (3)

where Ll_y = < Llu 1 , ...Llu 6 > T are the increments of nodal displacements and .!1. contains
derivatives of !f_, the standard linear and cubic shape functions for a beam element [5].
The force-incremental displacement relationship can be expressed as:

(4)

where the following summations over the fibers in the cross-section replace the integrals of
stress:

Po ~O'o;A;, Mo = ~O'o;A;y;

In the summations, E;, A;, y;, and u o; are the tangent modulus, area, distance from the
reference axis, and current stress of fiber i, respectively.
Upon use of Eqs. (3) and (4), the principle of virtual displacements gives the equi-
librium equations

JiJ Ll_y = Ll£. - (Eo - B.o) (5a)

where
Kt = J liT ll.fl.dz (5b)
392 Kreger, Fenves, and El-Habr

D.E = J!i_T flpdx (5c)

E.= Jlf...T E..dx (5d)

Eo= f IF { Jl. }dx (5e)

in which ill!. is the load increment from the current load l!.o· The quantity (Eo - Eo)
represents the unbalanced forces from previous load iucrcwcnts; it is zero if each load
increment is brought to equilibrium.
After solution for A!!., the strain in each fiber is computed from Eqs. (1) through
(3), the stresses are computed from the constitutive relationships, and summed to give P 0
and M •. Equation (5e) then gives the current internal resisting forces.

External Tendon Element

Neglecting slippage at the attachment points, the tendons are modeled by axial
elements connected through rigid links to the girder segments. The rigid links represent
the points of tendon attachment at diaphragms (or deviators) and anchorages, which, in
this analytical procedure, are assumed much stiffer than the box girder. Figure 2 shows
the geometry of the tendon element.
As with the beam element, the strain in the tendon can be represented by Eq. (1).
Assuming small displacements, the increment in strain is

(6)

where fl. = t p
< -1 1 > and A'Ji = < Au 1' Au 2' > in the local coordinates of the
tendon, and Lp is the length of the tendon. the force displacement relationship is obtained
by the principle of virtual displacements:

(7)

where D = =i,: Ep, F

p
o = Apu0 , Ap is the cross-sectional area of the tendon, Dp is the
tangent modulus of the tendon, and 0' 0 is the current stress in the tendon (which includes
prestress).
The displacement increment, A'Ji, at the ends of the tendon can be expressed in
terms of the displacement increments at the ends of the rigid links by the compatibility
relationship A!!.' = tA.Y, where A!!. = < Au 1, ... Au 6 > T and

t _
-
[c
0
s -f31
0 0
o o oJ
c s -{32
(8)

in which C = case, S = sine, {3 1 = e1C + g1 S and {3 2 = e2C + g 2 S (In this study

g1 = g2 = 0). Consequently, the force displacement relationship for the tendon element is:

(9)

The current strain in the tendon is given by A£ = fl. tAl!. and Eq. (1), which can be used
to evaluate the axial force in the tendon.
 Prestressing 393

Joint Element
Externally post-tensioned bridges have been constructed with no epoxy in joints
betwen precast segments. Under load, the joints may open, causing a redistribution of
internal resisting forces in the girders. The joint opening is modeled by assuming that
deflections are small, sections remain plane, and there is no slippage between segments
across the joint.
The element is shown in Fig. 3. The element has two rotational degrees and two
horizontal translational degrees of freedom to represent the opening of the joint. The ends
of adjacent beam elements (segments) are connected by a distributed spring that only
resists closing of the joint (Fig. 4a). The assumed kinematic relationship at the joint is
shown in Fig. 4b, where the gap is given by

g = .!l.!!. (10)

where .!l = < 1 - y - 1 y >, !!. = < u 1 , ••• u 4 > T, and negative g indicates opening of
the joint. The location, Yc, of the gap opening is defined by g = 0, or

(11)

The value of Yc, when compared to the values of g at the top and bottom, determines the
four states of the joint: fully closed, open at top, open at bottom, and fully open.
The force in the joint spring is I = kg over the depth of contact, s ~ Yc ~ r, where
s and r depend on the state of the joint opening, and I = 0 over the open portion of the
joint. Representing the force in the spring as I =
I:::. I + I 0 and the gap as g=l:::.g + g0 ,
the principle of virtual displacements gives the following force displacement relationship:
I

P. = lit f:::.y + /_o ( 12)

where P. = < Pl, ···P4 > T, l:::.y = < l:::.u1, ... l:::.u4 > T
and

!it = 1r .!lTk.!l dy (13a)

/_o = 1r .!lT lodY (13b)

Using the definition of .!land lo, closed-form expressions for lit and /_ 0 can be formed
in terms of the current state of opening of the joint (7]. The state of the joint and internal
resisting forces can be updated after solution for t::.y.
The joint stiffness, k, used in the model was 100,000 ksi. This was large in comparison
to the axial stiffness of the girder, but was not so large as to result in an ill-conditioned
stiffness matrix or oscillatory solution as determined from a parameter study. Additional
details on the formulation of the model and solution procedure can be found in Ref. 6 and
7.
394 Kreger, Fenves, and El-Habr

APPLICATION OF MODEL

The nonlinear finite element model described in the previous section was used to
analyze a three-span, post-tensioned segmental box-girder bridge with external tendons.
Comparisons are made between computed behavior of a girder with dry joints between
precast segments and behavior of a similar monolithic structure. In addition, behavior of
the segmental box girder with additional bonded internal tendons is examined.
During development of the finite element model and the course of this investigation a
reduced-scale, three-span, post-tensioned segmental box-girder bridge model with external
tendons was designed and fabricated at The University of Texas at Austin. As a matter of
convenience the basic details of the experimental model were used as input for the finite
element model. A plan view and elevation of the model are presented in Fig. 5, and the fiber
models for typical segments and pier segments are shown in Fig. 6. In the finite element
model each segment is divided into two elements, and closure strips at the end of each
span are considered as separate elements with the same cross section as typical segments,
but without reinforcement. The area of each fiber, distance from the fiber centroid to
the reference axis, and material types are given in Tables la and lb. Dead load for the
box-girder model is 1.80 kips/ft, and is applied as concentrated forces at nodes in the finite
element model.
Material behavior is idealized by a second-degree nonlinear stress-strain relationship
for concrete [4], and a bilinear curve for each type of reinforcing steel.
Table 2 lists the properties for the different material types. The external tendons
are divided into elements that span between anchorage locations and deviation points at
the diaphragms, and have a bilinear stress-strain relationship (Fig. 7).
I
The external tendon geometry and prestressing forces used in the finite element
analyses are presented in Fig. 8. Data for the tendons are tabulated below each span. Two
lines of data are presented for each tendon. Each entry of data in the top line corresponds
with the vertical distance of the tendon below the reference axis (Fig. 6) at deviator
locations. Data in the second line correspond with the tendon forces between deviators.
Tendon force is constant between deviator locations because tendons are not attached to
the girder between adjacent deviator locations or between anchorage and deviator locations.
Failure at a joint-opening location corresponds with crushing of concrete in the
compression flange. Concrete crushing is assumed to occur when an average stress of
0.85!' c acting over a rectangular area defined by the width of the compression flange and
a depth equal to fJ1c (as defined in ACI 318-83 [8]) is needed to satisfy equilibrium at a
joint.

Computed Behavior of Segmental Versus Monolithic Construction

Behavior of the segmental girder is compared to the behavior of a similar monolithic

girder where the reinforcement is assumed continuous except through the closure strips.
Both structures are subjected to a concentrated applied load on the interior midspan. The
load is incremented up to failure which is manifested by either the crushing of concrete at
a joint (limiting concrete stress is reached) or by the inability of the computer solution to
achieve equilibrium.
The segmental girder fails when the interior midspan joint opens to approximately
94% and the average concrete compressive stress exceeds its limiting value at the joint.
 Prestressing 395

However, the monolithic girder first forms a hinge at midspan when the corresponding
prestressed and non-prestressed reinforcement yields. Then, a mechanism develops when
two more hinges form at the closure strips of the interior span because concrete reaches its
limiting strength at the bottom fibers of the girder.
The applied load versus displacement response, and response of one of the tendons
at interior midspan for both girders are plotted in Fig. 9 and 10. The maximum load
capacity, deflection, and tendon stress are 56 kips, 0.67 in., and 194 ksi, respectively for
the segmental girder. In contrast, the maximum load capacity, deflection, and tendon stress
for the monolithic girder are 108 kips, 2.8 in., and 245 ksi.
The low strength and ductility of the externally prestressed segmental girder can be
attributed to the unbonded tendons and dry joints. Because the tendons between deviation
points are unbonded, the strain in the tendon is constant along its length, and the strain at
the critical section is less than what it would be for internally bonded tendons under similar
conditions. Hence, the stress in the tendons increases very little so that when the crushing
strain has been reached in the concrete, stress in the prestressing steel is far below its
ultimate strength (Fig. 10). Also, the existence of joints tends to concentrate compressive
stresses in the concrete at the joints, resulting in "premature" failure.
In the monolithic girder, multiple cracks form in the interior midspan. The absence
of the joints eliminates the problem of concentrated compressive stresses because the cracks
1
are not as deep as the joint openings in the segmental girder. Also, the increased size of
the compression zone allows stress in the tendons to increase to yield at an applied load
of 100 kips. Furthermore, the improved section behavior afforded by distributed cracking
also results in a large increase in ductility.
The degree of ductility reduction indicated for the segmental girder should be viewed
with some amount of caution because of potential errors associated with the assumptions
incorporated in the analysis of the opening joint near ultimate loads. When the joint-
opening width becomes large, it is unlikely that the joint section can be considered plane
any loanger. Also, any overestimation of joint stiffness increases the fraction of the joint
that opens at a given load, and results in a concentration of compressive stresses in the
flange that leads to joint failure at a reduced load.

Computed Behavior of Segmental Girder with Bonded Internal Tendons

The segmental girder is considered here with bonded internal tendons in the top
and bottom flanges of the segments. These tendons are composed of the same material
as the external tendons, with a total cross-sectional area of 0.68 sq. in. in the top flange
and 0.34 sq. in. in the bottom flange. The tendons are located at 1.3 in. from the
top and bottom of the section, and are stressed to 185 ksi (before elastic shortening). In
the analysis, the internal tendons are approximated by end forces and two additional steel
fibers in each segment. The response of the girder with internal tendons is compared to the
response of the segmental girder in Fig. 11 and 12. The addition of bonded internal tendons
results in an increase in capacity (from 56 to 65 kips), and a slight reduction in maximum
midspan deflection. The internal tendons increased the compression stress acting over the
entire girder cross section which slightly reduced the initial stress in external tendons at
midspan, and substantially reduced internal tendon stresses at ultimate (See Fig. 12). The
added internal tendons also reduced the width of opened joints.
396 Kreger, Fenves, and El-Habr

CONCLUSIONS

Finite element analysis of an externally post-tensioned segmental box girder yields

the following conclusions:
(1) The segmental box girder with external tendons and dry joints (no epoxy) between
segments exhibits reduced strength and ductility when compared with a similar
monolithic girder. The reduced strength and ductility are attributed to the unbonded
tended lengths and dry joints.
(2) The monolithic box girder develops multiple cracks in the midspan region. The
deformation capacity available at each crack location permits tendon deformations
between deviation points (location of tendon attachments) large enough to yield the
tendons.
(3) Supplemental bonded internal tendons improve flexural strength and slightly reduce
ductility. Behavior of the girder with supplemental bonded tendons indicates the
utility of keeping joints closed as long as possible.

ACKNOWLEDGEMENTS

This paper is based on research sponsored by the Texas State Department of High-
ways and Public Transportation and the Federal Highway Administration. All opinions
and conclusions expressed are those of the writers and do not necessarily represent the
views of the sponsors.
 Prestressing 397

REFERENCES

1. Kaba, S.A., and Mahin, S.A., "Refined Modelling of Reinforced Concrete Columns
for Seismic Analysis," Report No. UCB/EERC- 84/03, Earthquake Engineer-
ing Research Center, University of California, Betkeley, April, 1984.
2. Kang, Y.-J., and Scordelis, A.C., "Nonlinear Analysis of Prestressed Concrete
Frames," Journal of Structural Engineering, American Society of Civil En-
gineers, Vol. 106, No. ST2, Feb. 1980, pp. 445-462.
3. Mahasuverachai, M., and Powell, G.H., "Inelastic Analysis of Piping and Tubular
Structures," Report No. UCB/EERC- 82/27Earthquake Engineering Research
Center, University of California, Berkeley, Nov., 1982.
4. Hognestad, E., "A Study of Combined Bending and Axial Load in Reinforced Con-
crete Members," Bulletin No. 399, Engineering Experiment Station, University
of Illinois, Urbana, Nov., 1951.
5. Zienkiewicz, O.C., The Finite Element Method, 3rd Ed., McGraw-Hill, 1977.
6. Fenves, G.L., "Nonlinear Analysis of External Prestressed Bridges," Proceedings,
Ninth Conference on Electronic Computation, American Society of Civil Engineers,
Birmingham, Alabama, Feb. 1986, pp. 192-201.
7. El-Habr, Kamal C., "Finite Element Analysis of Externally Prestressed Segmental
Construction," unpublished M.S. Thesis, The University of Texas at Austin, May
1988.
8. American Concrete Institute Committee 318, "Building Code Requirements for Re-
inforced Concrete (ACI 318-83), American Concrete Institute, Detroit, Michigan.
398 Kreger, Fenves, and El-Habr

Table 1 Fiber Data

(a) Typical Segments

Fiber Area (in.') y (in.) Material*

1 84.00 -6.00 c
2 1.26 -5.50 S3
3 0.40 -5.50 S1
4 84.00 -5.00 c
5 54.00 -4.09 c
6 15.90 -2.78 c
7 11.70 -1.29 c
8 12.10 0.49 c
9 12.00 2.50 c
10 9.00 4.25 c
11 18.00 5.65 c
12 36.00 7.06 c
13 48.00 .8.00 c
14 0.40 8.50 S1
15 0.60 8.50 S3
16 48.00 9.00 c
See Tab! e 2
C = Concrete; S = Steel

(b) Pier Segments

Fiber Area (in.') y (in.) Material*

1 42.00 -6.00 c
2 84.00 -5.50 c
3 1.99 -5.44 S2
4 84.00 -4.50 c
5 0.85 -4.12 S1
6 84.00 -3.50 c
7 48.50 -2.25 c
8 61.70 -0.50 c
9 60.00 1.50 c
10 60.00 3.50 c
11 49.50 5.27 c
12 48.00 6.50 c
13 0.50 6.62 S1
14 48.00 7.50 c
15 48.00 8.50 c
16 1.98 8.71 S3
17 24.00 9.25 c
!:iee Tab e 2
C = Concrete; S = Steel
 Prestressing 399

Table 2 Material Properties for Concrete and Reinforcing Steel

Material Type* Description

c (e = 6.00 ksi (I = 562 ksi
E; = 5500 ksi Ceo = 2.20 x 10- 3
S1 f,y = 45.0 ksi c,u = 0.16
E,1 = 29000 ksi E,2 = 100 ksi
S2 f,y = 60.0 ksi c,u = 0.21
E,! = 29000 ksi E,2 = 100 ksi
S3 f,y = 80.0 ksi c,u = 0.28
E,1 = 29000 ksi E,2 = 100 ksi
See Table 1
Ceo =
ultimate concrete strain
c,u = ultimate steel strain
400 Kreger, Fenves, and El-Habr

0 Beam Element

0 Joint Element

D Tendon Element
I Rigid Link

Figure 1. Finite element model for a simple-span, segmental girder with

external tendons.

Tendon

e Lp
e2

Rigid
t u5
e1

tu2
"• (j. • u,
Links

"•( ~l
L
.I \

Figure 2. External tendon element

 Prestressing 401

u4
-
.. u1 _t Reference
Axis

Figure 3. Joint element model

f(Y)
Compression
Compression

y
Reference Axis

Tension g (Y)

Figure 4a. Gap between segments Figure 4b. Force-displacement

relationship for the joint
 ..t:;..
SYMMETRICAL_:_ 0
~~-
1'-:l

r-PIER SEGMENT rTVPICAL SPAN SEGMENT rc.I.P. CLOSURE STRIP I

I
~
C6
!I I: :I I: :I !I I: :I I: II I: :I !I I: :I I: ~
,.'"1

------ - ---- ------- ------ ------- ----·-- ------ ------- ------ ------ .................. ------- ------ ------ ------- ------· ~
:I I! ! I I: :I I, I I! I[ I: II J.lhlilli 1: i_l 11 Ji ! I I: ::s
~
Jl "
(Jl

ll:l
4'-0" BOTTOM FLANGE ::s
PLAN VIEW Q...
M
,_.
I

::I:
ll:l
0""
'"I

ELEVATION

Figure 5. Model dimensions

 Prestressing 403

I
L 1
2,3
l 4
5
6.5"

Reference Axis 'r-f

-
6
7

8
r-- I

9
y r-
10
r-
........... 11
L
___..,r 12
L 13
14,15
l 16

Figure 6a Typical segment fiber model

I
4

6.5" ~
j 7

Reference Axis I 8

10
y

"'
11
12
~~
15
16
17

Figure 6b. Pier segment fiber model

404 Kreger, Fenves, and El-Habr

300
ftu
fty
--------------------
I
I
I
200 I
I
I
I
I
I
I
100 I
I
I
I
.00838 = f:ty 1.0623 = etu

0.01 0.02 0.03 0.04 0.05 0.06 0.07 £

rn
Figure 7. Stress-strain relationship for external tendons

·:"fl J£0 11111111111r= 111111111111111 7Jk ~

Span A T•"""• """""'

Tendon A1 : 10-318" dia., A ps = 0.85 in~

.....
y (in.) ~ 3.99 5.75 6.33 6.33 5.75 -.77 "'~
p (kips) 161 161 157 153 149 144 138

Tendon A2. · 10.318" dia., A ps= 085 in 2

0 ;:::
y (in.) ~ -.87 5.75 6.04 6.04 5.75 -1.48~

p (kips) 160 159 154 150 146 141 134

Tendon BA: 4-318" dia., A ps = 0.34 in~

co
0 .....
y(in.) ~ -1.15 5.75 5.75 -1.84~

p (kips) 64.0 63.6 62.5 61.3 60.4

Figure 8. Tendon geometry and prestressing forces

 Prestressing 405

*
Span B

~ lllllllllllllllllllllllllllllll
:;& ;:JJ;;
Tendon B: 10-3/8" dia, A ps =0.85 in.2
...,.
~
y (in.) co
-.82 5.75 6.33 6.33 5.75 -.82 ~

P(kips) 119 126 131 134 138 144 153

Tendon BA: 4-318" dia., A ps = 0.34 in~

CD
CD
..... .....
y (in.) ~-1.54 5.75 6.04 6.04 5.75 -1.54~

P(kips) 59.6 58.7 56.3 54.8 53.5 51.2 48.7

Tendon CB 4-3/8" d"1a, A ~= 034.ln.2

It)
It)
CD CD
y (in.) ~ -1.91 5.75 5.75 -1.91 ~

p (kips) 53.9 56.3 57.4 58.7 59.5

Span C

~ lilllllllillillllllllllllllllll ;:y; I
Jh7
Tendon C1 : 10-3/8" dia., A ps =0.85 in~
..... It)
y (in.) ~ -.77 5.75 6.33 6.33 5.75 3.99<'i
p (kips) 138 144 149 153 157 161 161

-
Tendon C2 ·10 3/8" dia., A ps=
0
y(in.) ~ -1.48 5.75 6.04 6.04 5.75 -.87 ~

P(kips) 134 141 146 150 154 159 160

Tendon CB: 4-3/8" dia., A ps = 0.34 in~

~ 0
y(in.) ~-1.91 5.75 5.75 -1.15 ~

P(kips) 60.4 61.2 62.5 63.6 64.0

Figure 8. Tendon geometry and prestressing forces

(continued}
406 Kreger, Fenves, and El-Habr

120

100
0 Monolithic
0.
;g. 80
i
.3 60
j
a.
0. 40
<(

20

0
0 1 2 3
Displacement (in.}

Figure 9. Applied load-displacement curve at interior midspan

260

240
--
·u;
.::J!
220
0
0
e
us 200
c:
0
"0
1:: 180
~
160

140
0 20 40 60 80 100 120
Applied Load (kips}
Figure 10. Applied load versus interior midspan stress of tendon B
 Prestressing 407

190 .---------------------------~r-----------~

180
"iii
e. Without Internal
If)
If) Tendons
~
Ci5 170
c:
0
"0
c:
Cl) W~h Internal
1- Tendons
160

150 +-----r---~-----r----~----~--~-----r--~
0 20 40 60 80

Applied Load (kips)

Figure 11. Comparison of the load-deflection response for segmental gird-

ers with and without internal tendons

80

Ill 60 With Internal

0. Tendons
~
-g
0
..J 40
-o
Cl> WithOut Internal
~ Tendons
0.
<t: 20

0.0 0.2 0.4 0.6 0.8

Displacement (in.)
Figure 12. Comparison of tendon stress-applied load response for seg-
mental girders with and without internal tendons
 SP 120-20

Computer Evaluation of the Effect of

Prestressing on a Shell Structure
by J.M. Nagele, S.C. Das, and R.M. Bakeer

Synopsis: The purpose of this investigation is to ex-

amine the forces developing in the dome of a
prestressed concrete nuclear containment structure due
to prestressing of vertical tendons. A computer
program was developed for personal computers to
evaluate the effect of prestressing forces. A
parametric study was conducted, using the program, to
examine the effect of dimensions, prestressing force,
properties of tendons, etc. on the resulting
stresses in the dome.

Keywords: oonputer programs; domes (stnlctural fonns);

evaluation; jack.irx.J; nuclear reactor contairnnent; prestressed
concrete; prestressing; prestressing steels; shells (structural
fonns); structural analysis

409
410 Nagele, Das, and Bakeer

J.M. Nagele is a former graduate student at the Civil

Engineering Department, Tulane University. She is
currently a Design and Construction Engineer at
Chevron USA.
S.C. Das and R.M. Bakeer are Associate and Assistant
Professors of Civil Engineering at Tulane University.
J.M. Nagele and R.M. Bakeer are Associate Members of
ASCE and S.C. Das is a Member of ASCE. R.M. Bakeer
is an ACI Member.

INTRODUCTION

A program was developed for personal computers to

determine the forces induced by the prestressing of
vertical tendons of the dome of a concrete nuclear
containment structure. The structure to be analyzed
consists of a cylinder topped with a hemispherical
dome. A total of 42 mutually perpendicular, two-way,
U-shaped, vertical prestressing tendons are anchored
below the foundation slab of the structure in tendon
galleries, as shown in Fig. 1. These tendons run ver-
tically through the cylindrical wall and continue
through the dome down to the other side of the wall
where these tendons are also anchored. The tendons
form a series of adjacent semicircles over the dome
with the largest semicircle near the apex of the dome.
Fig. 1 shows a cross-section of the structure under
consideration, and Fig. 2 shows the plan view of the
arrangement of the 42 U-shaped vertical tendons. This
particular arrangement is economical because it intro-
duces the prestress in the dome and wall simul-
taneously and therefore reduces the number of required
anchorages. Even though horizontal tendons were also
used in the cylindrical part of the structure, the
focus of this paper is on the forces in the dome part
of the structure and only the effect of prestressing
of vertical tendons is discussed herein. The computer
program [2, 5] was developed to determine the forces
in the vertical tendons while considering curvature
and friction losses. Losses due to length and wobble
effects are neglected in this study because these
losses are relatively small in magnitude when compared
with friction losses [2]. Therefore, the prestressing
losses of the tendons in the cylinder are neglected,
and only friction losses in the dome are considered.

THE COMPUTER PROGRAM

The computer program is written in FORTRAN 77

for use on IBM type personal computers with 640 KB
memory. The program is compiled using the IBM Profes-
 Prestressing 411

sional FORTRAN Compiler [8]. A complete listing of

the main program, its subroutines and input/output
procedures are given in Ref. [5].

The prestressing of tendons in a prestressed con-

crete shell yields normal pressures acting toward the
center of the dome and frictional forces acting along
its curved surface. Fig. 3a shows the radial forces
acting on the dome, whereas Fig. 3b shows the forces
due to curvature and the change in direction of these
forces at an angle a'. The program computes the nor-
mal and frictional forces in terms of cylindrical
coordinates (r, a, z) and then translates these forces
into spherical coordinates (R, ~. e). The three com-
ponents of pressure in spherical coordinates are the
radial, circumferential and meridional pressures. The
radial pressures act normal to the curved surface,
whereas the circumferential and meridional pressures
act in the plane of the curved surface. However, the
circumferential pressures are in the direction perpen-
dicular to the tendons, and the meridional pressures
are in the direction parallel to the tendons.

ANALYSIS OF THE DOME FORCES

The dome is divided into several regions defined

by the user in the input for the computer program.
Due to symmetrical vertical tendon layout in this par-
ticular structure, only one-fourth of the hemispheri-
cal dome needs to be analyzed. The section considered
in the analysis, equivalent to a planar sector of
circle with an angle of n/2, is divided into horizon-
tal rings and vertical slices, as shown in Fig. 4. To
facilitate direct numerical integration in the
program, the segment length of a region is chosen to
be less than {Rt, where R and t are the minimum
average radius and the thickness of the shell, respec-
tively. The quadrilaterals formed by these intersect-
ing lines are defined as regions. These regions are
numbered from the starting point, where the azimuth
angle, e, equals 0.0, and the colatitude angle, ~. is
equal to n/2. The numbering proceeds to values of e
of n/2 and ~ of n/2 and up to ~. equal to 0.0.
However, the regions within the two rings at the apex
are considered as one because the area of these in-
dividual regions becomes infinitesimal. The program
determines the angular boundaries of each region and
calculates the corresponding areas [2].

The friction forces are calculated as well as the

normal forces for increments along the curved vertical
tendons. The radial, circumferential, and meridional
pressures are then calculated and transformed into
terms in a Fourier series. With the pressure on the
412 Nagele, Das, and Bakeer

shell described as a Fourier series, a finite element

program, such as Kalnin's "Static Analysis of Thin
Elastic Shells of Revolution" [9], can be used for the
computer analysis. Detailed description of the mathe-
matical calculations involved in the program are
available in the Ref. [2].

SELECTION OF THE PROGRAM INPUT

The computer program was originally developed to

calculate the forces in a specific shell. The charac-
teristics of that shell are used in this study as the
control parameters. Several executions of the program
were conducted to investigate the effect of various
parameters on the resulting prestressing forces in the
dome. In all cases, only one parameter was changed at
a time with the remaining parameters defined as their
control values. Table 1 shows the different values of
the various parameters investigated in this study as
well as their control values, (bold values in
parentheses). The upper and lower bounds of the
parameters shown in Table 1 are selected according to
the guidelines provided in Ref. [1, 3, 4, 6, 7]. The
number of Fourier harmonics was not varied, because
this parameter has no effect on the calculation of the
pressures due to prestressing.

RESULTS OF THE ANALYSIS

Selected regions in the dome are chosen at

various locations on the grid to demonstrate the
results of the analysis. These regions are numbers
11, 36, 50, 75, and 107 shown in Fig. 4. Variations
in the values of radius, friction coefficient, ul-
timate strength, number of tendons, coefficient of
jacking stress and coefficient of lock-off stress ver-
sus radial pressure were plotted on the semi-
logarithmic graphs shown in Fig. 5 through 8. The
colatitude and azimuth pressures on regions 11 and 107
are tabulated in Tables 2 and 3 for the various
parameters under investigation. Tables 1, 2 and 3 are
interrelated where Table 1 shows the values of the
different parameters under investigation, whereas
Tables 2 and 3 show the effect of varying one par-
ticular parameter at a time on the resulting pressures
in the dome. Columns 2 through 5 in Tables 2 and 3
show the resulting pressures in the dome when using
the four values of the indicated parameter in the
first column along with the control values of the
other parameters. For example, the first row of Table
2 shows the circumferential pressures in region 11 of
the dome for radii of 50, 60, 72 and 85 feet, using 42
tendons with ultimate strength of 2230 ksf, a friction
coefficient of 0.3, a coefficient of jacking stress of
 Prestressing 413

0.85, a coefficient of lock-off stress of 0.7, 5

Fourier harmonics and a number of divisions of 9 and
14 in the azimuth and colatitude directions, respec-
tively.

In all cases, the effect of the changing

parameter on the radial pressure was similar to the
effect on the circumferential and meridional pres-
sures. The results shown in Fig. 5 and Tables 2 and 3
show that by increasing the radius of the dome, the
radial pressures decrease. This is true because the
increase in radius increases the areas of the in-
dividual regions and the force on each region is then
distributed over a larger area resulting into lower
pressures.

The friction coefficient has a greater effect on

the regions near the apex than the other parts of the
dome, as shown in Fig. 6. The lower regions, numbers
11 and 36, show a negligible variation in radial pres-
sure with the variation in the friction coefficient.
However, for the upper regions near the apex, the
pressure decreases with increasing the friction coef-
ficient. This decrease in pressure is due to the
greater frictional losses in prestress forces along
the curved surface of the dome.

Fig. 7 demonstrates that when using tendons with

higher ultimate strength, the induced pressures in-
crease because, for greater strengths, the tendons are
stressed more.

The effect of the number of tendons on the radial

pressures is presented in Fig. 8. Although there is
some random scatter of the data, the best fit lines
are positively sloping. The graph shows that by in-
creasing the number of tendons, the induced force in a
given region increases. For one stressed tendon pass-
ing through a particular region, a force is induced in
that region. On the other hand, if two stressed ten-
dons pass through a region, then the force induced in
that region doubles. A direct comparison can be made
from the resulting pressures for the same dome when
using either 25 and 50 tendons. The radial pressure
on region 11 for 25 tendons is 5.3 ksf while the
radial pressure for 50 tendons is 10.6 ksf for a
direct ratio of 1 to 2.

The coefficients of maximum jacking stress and

lock-off stress have only a minor effect on the forces
induced by prestressing. There is no more than a 15
percent increase in the pressures at the different
regions due to changing the coefficient of jacking
414 Nagele, Das, and Bakeer

stress from 0.75 to 0.90. For the increase in coeffi-

cient of lock-off stress from 0.65 to 0.80, the in-
crease in pressure is less than 14 percent.

The number of divisions in the azimuth angle

greatly affects the pressures. When the dome is
divided into 4 sections in the azimuth direction,
region 18 encompasses a given area. Regions 35 and 36
of the same dome divided into 8 sections includes the
same area as region 18 of the 4 sectioned dome.
Likewise, regions 52, 53, and 54 of the same dome
divided into 12 azimuth divisions also describe the
same area of the dome surface. The resulting pres-
sures in these regions are listed in Table 4. The
pressures in regions 35 and 36 are, in general,
greater than the pressure in region 18. Similarly,
the pressures in regions 52, 53, and 54 are nearly
triple the pressure in region 18. This is true be-
cause the same force is applied to the given area
identified by these regions. The force is then
divided into two or three regions and then divided by
the individual areas. Thus, ideally the same pressure
would be applied to this given area defined by mul-
tiple regions. Therefore, the pressure in a given
area is directly proportional to the number of regions
into which the area is broken.

As before, selected regions occupying the same

area are chosen for comparison. Region 20 of the dome
divided into 6 colatitude sections, regions 38 and 47
of the dome divided into 12 colatitude sections, and
regions 56, 65, and 74 of the dome divided into 18
colatitude sections all describe the identical area of
the dome surface. The azimuth direction is defined by
angle e and the colatitude direction is defined by the
angle ~. as shown in Fig. 4. Table 5 presents the
pressures induced in these regions. A direct
relationship cannot be identified as easily as with
the variations in azimuth divisions. Dividing a given
area by two increases the total pressure in that area,
However, the ratio of azimuth divisions to pressure is
not consistent due to the curvature of the dome and
the resulting friction losses. Dividing the region
into smaller areas describes the curvature of the dome
more accurately as the smaller regions would more ac-
curately define the pressures in that area.

CONCLUSIONS

The results of the parametric study highlight the

relationship between the different input variables and
the pressures produced by prestressing of the vertical
tendons. For each region the resultant prestressing
 Prestressing 415

force was identified by forces in the radial, circum-

ferential, and meridional directions. Parameters
which increased pressure on the regions include in-
creasing ultimate tendon strength, number of tendons,
coefficient of jacking stress, and coefficient of
lock-off stress. A decrease in pressure resulted when
the radius increased or the coefficient of friction
increased. The results of the program for various
azimuth and colatitude angle divisions showed that the
divisions in the colatitude angle are more critical.
In the analysis of a dome, it should be divided into a
large number of colatitude rings to ensure a greater
accuracy. The friction losses would then be more ac-
curately defined, and thus, a more precise output of
pressures would result. The results of the program
are useful in predicting the outcome of the pressures
of the tendons of a prestressed concrete containment
shell. During design and construction phases of these
structures, important changes can be made. The com-
puter program provides a quick method of determining
the pressures applied to a shell taking into account
the various parameters such as the radius, friction
coefficient, ultimate tendon strength, number of ten-
dons, coefficient of jacking stress, coefficient of
lock-off stress, divisions in colatitude angle, and
divisions in azimuth angle.

ACKNOWLEDGMENT

Part of the work described in this paper in con-

nection with the nuclear containment structure was
performed by the second author in conjunction with
other engineers at the firm of Brown & Root, Inc.,
Houston, Texas during the development stage of the
South Texas Project in 1972.

REFERENCES

1. Ashar, H. and Naus D.J., Overview of the Use

of Prestressed Concrete in U.S. Nuclear Power
Plants", Nuclear Engineering and Design, V. 75,
No. 3, June 1983, pp. 425-437.

2. Das, S.C. and R.M. Bakeer, "Analysis of the PWR

Containment Shell Structure Due to the Prestress-
ing Tendon Force", Paper Presented at the ACI
Fall Convention, Baltimore, MD, Nov., 1986.
3. Karwat, H., "Containments- Design Principles and
Background", Nuclear Engineering and Design, V.
90, No. 4, Aug. 1985, pp. 113-133.

4. Murray, D.W., "Observations on Analysis, Testing

and Failure of Prestressed Concrete
416 Nagele, Das, and Bakeer

Containments", Nuclear Engineering and Design,

V. 82, No. 1, Oct. 1984, pp. 37-46.

5. Nagele, J.M. •----~U~s~e~r-'~s~M~a~n~u~a~l~fwo~rL-t~h~e~T~U~E~S~4~2~0

Program, Tulane University, New Orleans,
Louisiana, 1988.

6. Priestley, M.J., "Analysis and Design of Circular

Prestressed Concrete Storage Tanks", Prestressed
Concrete Institute Journal, V. 30, No. 4, July-
Aug. 1985, PP• 64-85.

7. Shunmugavel, P. and Gurbuz 0., "An Evaluation of

Structural Failure Modes for Prestressed Concrete
Containments", Nuclear Engineering and Design, V.
104, No. 3, Nov. 1987, PP• 349-355.

8. "IBM Personal Computer Professional

FORTRAN", Ryan-McFarland Corporation, 1984.

9. KALNINS, A., "Analysis of Shells of Revolution

Subjected to Symmetrical and Nonsymmetrical
Loads", Journal of Applied Mechanics, Trans.
ASHE, Sep. 1964, pp. 467-476.
 Prestressing 417

TABLE 1--PARAMETERS CONSIDERED IN THE STUDY

Parameter Value

Radius of Dome (feet) 50 60 (72) 85

Friction Coeff, 0.08 0.15 (0.3) 0.35
Ult. Tendon Strength (ksf) 1000 1500 (2230) 2500
No. of Tendons 25 (42) 50 60
Coeff, of Jacking Stress 0.75 0.80 (0.85) 0.90
Coeff, of Lock-off Stress 0.65 (0.7) 0.75 0.80
Div. in Azimuth Angle 4 8 (9) 12
Div. in Colatitude Angle 6 12 ( 14) 18
No. of Fourier Harmonics ( 5) 5 5 5

(xxx) Control Value of Parameter

TABLE 2--PRESSURES IN REGION 11

PARAMETER CHANGED CIRCUMFERENTIAL PRESSURES (ksf)

Radius* -5.48 -3.80 -2.64 -1.90

Friction Coeff. -0.60 -1.23 -2.64 -3.13
Ult. Strength -1.19 -1.78 -2.64 -2.96
Number of Tendons -1.58 -2.64 -3.16 -3.70
Coeff, Jacking Stress 2.78 0.11 -2.64 -2.64
Coeff, Lock-off -2.45 -2.64 0.12 3.07

PARAMETER CHANGED MERIDIONAL PRESSURES (ksf)

Radius* 4.97 3.45 2.40 1. 72

Friction Coeff, 2.22 2.27 2.40 2.44
Ult, Strength 1.08 1. 61 2.40 2.69
Number of Tendons 1. 51 2.40 2.96 3.41
Coeff, Jacking Stress 2.06 2.22 2.40 2.40
Coeff, Lock-off 2.23 2.40 2.38 2.27
* Only Radius is changed and other parameters have
their control values.
418 Nagele, Das, and Bakeer

TABLE 3--PRESSURES IN REGION 107

PARAMETER CHANGED CIRCUMFERENTIAL PRESSURES (ksf)

Radius* 7.97 5.54 3.84 2.76

Friction Coeff. 2.82 3.26 3.84 3.93
Ult. Strength 1. 72 2.56 3.84 4.31
Number of Tendons 2.28 3.84 5.54 7.04
Coeff. Jacking Stress 3.47 3.59 3.84 4.11
Coeff. Lock-off 3.82 3.84 3.85 3.84

PARAMETER CHANGED MERIDIONAL PRESSURES (ksf)

Radius* -2.21 -1.54 -1.07 -0.77

Friction Coeff. -0.64 -0.83 -1.07 -1.10
Ult. Strength -0.48 -0.72 -1.07 -1.20
Number of Tendons -0.51 -1.07 -2.01 -2.58
Coeff. Jacking Stress -0.96 -1.00 -1.07 -1.14
Coeff. Lock-off -1.06 -1.07 -1.07 -1.07

* Only Radius is changed and other parameters have

their control values.

TABLE 4--PRESSURES FOR VARIATION IN AZIMUTH DIVISIONS

Pressure (ksf)
Divisions Region(s)
Radial Colatitude Azimuth

4 18 -11.36 6.9 1. 06

8 35 -7.48 3.68 2.27

36 -14.88 10.11 -0.15

12 52 -7.38 3.05 2.52

53 -11.57 7.22 1. 39
54 -15.12 10.42 -0.71
 Prestressing 419

TABLE 5--PRESSURES FOR VARIATION IN COLATITUDE

DIVISIONS

Pressure (ksf)
Divisions Region(s)
Radial Colatitude Azimuth

6 20 -7.79 3.10 1.07

12 38 -6.69 2.36 1. 22
47 -11.12 5.66 1. 21

18 56 -8.07 2.75 1.49

65 -6.30 2.13 1.00
74 -12.26 6.64 1. 21

CENTER UNE

SPRING UNE

'lc 'b

Fig. 1--Cross-section of Prestressed Concrete

Nuclear Containment Structure.
420 Nagele, Das, and Bakeer

21 Tenttnn,.

e- n/2

21 Tendons

e- n/4

e- = 0

Fig. 2--Plan of Vertical U-Shaped Tendon Layout.

c Tempora.,. jacldq lrtren

.gc 'i. f ••
"'
I-

E
(a)
~
"' c
l..
+'
(/)
2f
.
' '===~--:

... a

Center Line

(b)

Figure 3. Variation in Tendon Stresses.

 Prestressing 421

¢
)
....
.....

B = 0 e- = TI/2
¢ = TI/2 ¢ = TI/2

• Azimuth direction is defined by angle B

Colatitude direction is defined by angle ¢

Fig. 4--Identification of Regions.

~0

~
'\ l\ ~~
80 1\

~
\\
1\ \
70

60
1\~\
50
\;~ o
LEGEND

REGION 11
A REGION 36
D REGION 50
o REGION 75
e REGION 107
40

z 3 4 5 6 7 8 910 zo 30 40 :10 60 70 80 90 100

Ro.dlo.l Pr-e""ur'" (-!)W(kst")

Fig. 5--The Effect of Radius on Radial Pressure.

422 Nagele, Das, and Bakeer

0.40

I~

+'
c
0.30 1
Ill
u
<;::
""
Ill
0
u
o.eo
c
0
.p
u
I
u..
0.10
LEGEND
II 0 REGION 11
A REGION 36
D REGION 50
0 REGION 75

o.oo • REGION 107

2 3 4 5678910 20 30 40 50 60 80 100

Ro.cllo.l Pressure (-l)liE(ksf')

Fig. 6--Effect of Friction on Radial Pressure.

3000

2!100

ll
"Ill0. 1/
~ Ill
.<: 2000
+'
jl
/
0)
c:
Ill
1..
+'
V)
a, 1500
;I .L
+'
E
d
v/;~/
.;:;
:5
1000
/I; 'i'l/
LEGEND
0 REGION II
A REGION 36
D REGION :50
0 REGION 75
e REGION 107

2 "" 5 ' 7 8 910 20 30 .a 50 60 70 8090 100

Pressure (-l)liE(ksf)

Fig. 7--Effect of Ultimate Tendon Strength on Radial

Pressure.
 Prestressing 423

80

70

Ul
c 60
0
""0
c
QJ
I-
'+- 50
If ~11
v~~
0
!..
QJ
.Q
E 40 j
:::l
z

30
I; I/
vI/ I/
v f LEGEND

0 REGION II

vII
A REGION 36

~p{jl
0 REGION 50
o REGION 75
e REGION 107
20
2 3 4 5 6 7 8 910 20 30 40 50 60 80 100

Ro.dlo.l Pressure <-D•<ksf)

Fig. 8--Effect of Number of Tendons on Radial

Pressure.
 SP 120-21

External Versus Internal Bonded Prestressing

by E. Woelfel

Synopsis: In the near future prestressed concrete

structures will in Europe be designed according to
Eurocode 2 (EC2). The EC 2 principles governing the
design of structures with bonded tendons and with
external tendons are given in this paper, and a com-
parison is made between the structural behaviour of
the two different types of structures. It is shown
that the reliability of both is comparabl~ if the
characteristic strain due to prestressing (E t + s)
is introduced in the calculations. px -

Kevwords: building codes; limit state design; prestressed

concrete; prestressing; safety factor; serviceability; stress-
strain relationships; structural design; unborxled prestressi.rg

425
426 Woelfel

Eilhard Woelfel is Dr. -Ing. from the Technical Uni-

versity Aachen. He is Head of the Department of the
"Institut fuer Bautechnik" in Berlin (FRG). This In-
stitut has the role of coordination in the field of
Civil Engineering in Germany and takes part in these
efforts in Europe.

The Eurocodes

The Eurocodes are being drafted to serve in the

European Community as the officially recognized means
to prove the compliance of buildings and civil engi-
neering work with the legal requirements for structur-
al safety and serviceability. They shall promote the
functioning of the Internal Market by removing obsta-
cles rising from differing rules for planning and
design. A total set of 9 volumes is envisaged for the
Eurocodes.

Eurocode 1 - Common unified rules for different types

of construction and material;
Eurocode 2 - Concrete structures;
Eurocode 3 - Steel structures;

Eurocode 4 - Composite steel and concrete structures;

Eurocode 5 - Timber structures;

Eurocode 6 - Masonry structures;

Eurocode 7 - Foundations;

Eurocode 8 - Structures in seismis zones;

Eurocode 9 - Actions (design loads) on structures.

Eurocode 2 [1] gives the principles for reinforced

and prestressed concrete and the rules for application
for building only.

Rules for application for bridges will be elabo-

rated later on in part 2. This section will also con-
tain the design rules for external tendons.

For the moment, the author can only give his per-
sonal view on how to proceed with the design of struc-
tures with external tendons on the basis of the Prin-
ciples of EC 2. As a matter of the fact, some propo-
sals will be amended during the ongoing discussion,
 Prestressing 427

and some figures will change. This paper should be a

basis for these discussions, in the ACI-meeting and
later on in the EEC-Commissions.

Safety concept in Eurocode 2

The design equations in EC 2 are written in a form:

lp • S = R (fs/ls' fc/lc)
The partial safety factors lp and 1M are derived
from statistical considerations and are influenced
from:

the limit state considered

(higher safety factors for Ultimate Limit State
than for Serviceability Limit State)

the scatter of the parameters

(higher standard deviations lead to higher safety
factors)

the statistical definition of the parameters

(mean value or p-percent fractile)

compromises, which have to be made in order to

have a relatively small set of different factors.

The following limit states are considered:

Serviceability Limit State (SLS), which comprises

checking of:
crack width limitation
deflection
vibration
opening of joints of segmental construction
stresses in the prestressing steel

Concrete Stresses need not to be checked in SLS

because serviceability is not affected by the
stress in the concrete, provided that sufficient
safety against failure in the ultimate limit state
is provided for.

Ultimate Limit State (ULS) with the checks of:

normal forces (including stability)
bending moments
shear, torsion, punching
fatigue (requires special safety considera-
tions)

The parameters influencing the relevant limit

428 Woelfel

states are described by:

their characteristic (reference) value which is

the mean value of parameters whose scatter is re-
latively small (e.g. selfweight) or a certain
fractile in the other cases (usually 5 % for mate-
rial properties and 99 % for actions)

their standard deviations respectively the varia-

tion coefficient v.

Regarding prestressing effects, two characteristic

values are mentioned in EC 2: P and P which (ac-
cording to the author's mind) a~~uestima~~~ as + 17 %-
fractiles (P + 1.0 • s for normal distributions).
Reliability ~f&8ies using this characteristic value of
prestress show that sufficient reliability can be
achieved by applying the safety factor for prestress-
ing equal to unity [4].

The set of safety factors for the SLS and the ULS
is given in tables 1 and 2. The values taken from re-
liability studies [4] are shown in brackets for demon-
stration. When comparing the safety factors for pre-
stressing in ULS, one has to take into account that
the reference value in EC 2 is the mean value whereas
the author uses the 17 %-fractile as reference value.
The factors ~F (for self weight and live load) and lM
contain a partial coefficient taking account of the
uncertainty of the calculation which (in ULS) is
chosen as 1.05.

The material safety factors (table 2) are identi-

cal regardless of the limit state.

Eurocode 2 part 1 deals with concrete structures

only as far as they are prestressed by bonded tendons.
Structures with unbonded tendons (internal and exter-
nal) shall be treated in part 2 of EC 2 which will be
prepared on the basis of the principles given in EC 2
part 1.

The question is whether the same so.fety elements

should be chosen for both bonded and unbonded tendons.
Before this question will be answered, the main dif-
ferences in the design between bonded and unhanded
tendons shall be discussed.

The treatment of prestress in the design calculation

In the following section, the effects of pre-

stressing shall be treated first on the basis of
 Prestressing 429

strains which are induced in the steel by tensioning

of the tendons and which change locally due to fric-
tion and time due to time dependant strain variations
(see fig. 1). At the timet= 0 and at the active end
of the tendon x = 0, the tendon is strained to
E • The real value will not deviate very much
fr8£•I¥:'h~ calculated value, the standard deviation is
small and can be disregarded.

At t = t and at any point x of the tendon, the

steel strain will decrease by LlEp due to friction,
creep, shrinkage and relaxation. These strain losses
scatter considerably, and it is difficult to predict
them. Estimated standard deviations of these strain
losses are given in table 3 [ 2] (3]. Taking account of
the higher and lower fractile of the strain, the upper
and lower fractile of the stress and thus of the ten-
don force in SLS can be read from the ·-stress-strain-
li_!l~ of the prestressing steeT.

Going from SLS to ULS, a strain increase due to

the load increase will be observed in the tendon. Re-
garding a certain section of a structure with bonded
tendons, the strain increase of the tendon Ll£ 0 cross-
ing the crack is considerable ( ~ 10 %o ace. t:o EC 2)
and is to be added to the strain in SLS. Going from
the upper and lower value of the strain into the
stress-strain diagram, the corresponding stress is
(for both strain values) f k/ 0 . The ultimate capacity
of statically determinate Ystrfictural elements is - as
a rule for elements with bonded tendons - not influ-
enced by the scatter of the prestress.

Regarding the whole span of a structural element

(e.g. in order to calculate the statically indetermi-
nate part of the prestressing action), the medium
strain increase over the length ~Ll£px/L is consider-
ably smaller than the strain increase at a discrete
section. The overall value of the strain will there-
fore not reach the yield strain of the steel; the sta-
tically indeterminate part of the prestressing action
is therefore influenced by the scatter of the steel
strain.

In structural elements with unhanded tendons, the

strain increase of the tendons is due to the inte-
grated deformation of the structure between the an-
chorages of the tendon and is therefore small. The
overall strain in ULS will not reach the yield point.
Thus, the stresses in unhanded tendons in ULS scatter
in the same sense as in SLS.
430 Woelfel

The design of structures with unbonded tendons in

ULS must take account of the uncertainties of the pre-
stressing force whereas in structures with bonded ten-
dons in ULS, only the statically indeterminate part of
the prestressing action is influenced.

This difference in the mechanical behaviour of the

structure has major consequences: whereas the level
and the scatter of the prestress do not have a big in-
fluence on the reliability of a structure with bonded
tendons in ULS, this influence for structures with un-
handed or external tendons is very great. Thus, the
engineer designing a structure with unbonded tendons
must be interested in:

reducing the prestressing losses,

reducing the scatter of the losses respective to

the increase reliability of their prediction

designing rules such that these efforts lead to a

reduction of costs.

Under these points of view, it shall be discussed

whether the design rules in EC 2 are also adequate for
structures with unbonded external tendons.

Possible variation of the tendon force

The variations of the tendon force will be influ-

enced by:

the type of the structural element

(e.g. bridge deck or beam)

the tendon profile

the scatter of the mechanical parameters with in-

fluence on the prestressing losses (e.g. friction
or creep coefficient)

For the comparison between internal bonded and

external tendons, a special example of a beam will be
regarded with adequate tendon layouts. For this exam-
ple, it shall be shown whether fundamental differences
between the stress variations of bonded and external
unhanded tendons exist or not.

The main parameters of this example are:

tendon length 80 m
 Prestressing 431

distance between the section regarded

and the active end of the tendon
(one end stressing) 50 m

sum of deviation angles 35 °

unintentional angle of deviation
(only internal tendons) 50 • 0.005 15 °
mean concrete strain due to creep at
the level of the tendon

for an internal tendon 0.7 • 10- 3

for an external tendon 0.6 • 10- 3

The expected value of the strain losses are calcu-

lated on the basis of these assumptions (see table 3).

The comparison shows:

The losses of prestrain of an internal tendon are

about 1.5 times higher than for an external ten-
don. This is mainly due to higher friction losses
caused by a higher friction coefficient and unin-
tentional angles of deviation of the cable layout

The standard deviation of the losses is also about

1.5 times higher for internal bonded tendons than
for external tendons

The variation coefficient v of the prestrain re-

lated to the overall strain is the quotient:
s s
v (E) Epxt Epo,max - ~Ep

With the figures of table 3, the results for the

example are:

v 8.9 % for internal, bonded tendons

v 3.8 % for external tendons

Since the assumptions for the comparison have been

chosen rather conservatively, it may be generally as-
sumed:

v ~ 10 % for internal, bonded tendons

v ~ 5 % for external tendons.

432 Woelfel

Recommendations for the incorporation of the prestress

in the design

In the Serviceability Limit State and in the Ulti-

mate Limit State, the characteristic values of the ten-
don force shall be calculated from the stress-strain
curve using an upper and lower fractile of the strain
in the tendon. The safety factor on8£ is 1.0.
p
The fractiles of the tendon strain may be assumed
to be:

Epk £ pxt + 1.0 . s

1.1
0,9 } Epxt for internal
tendons
1. 05} £ px t for external
0,95
tendons

The stress-strain curve is limited to a value

fyk/ l s·

Literature

1 Eurocode 2: Common Unified Rules for Concrete-

Reinforced and Prestressed Concrete-Structures.
Commission of the European Communities, Brus-
sels Draft 1984

2 CEB-FIP Model Code 1990, First Predraft 1988.

Comite Euro-International du Beton: Bulletin
d'Information 190, July 1988

3 FIP State of the Art-Report: Tensioning of Ten-

dons. Force-Elongation-Relationship. Th. Tel-
ford Ltd. London 1986; ISBN 0 7277 0260 2

4 Hosser, D.; Koenig, G.; Theile, V.: Comment to

the Safety-Regulations in Eurocode 2 (in
German). Unpublished report. Institut fuer Bau-
technik Berlin 1985
 Prestressing 433

Notation A area
L span
p tendon-force
R resistance (bearing capacity)
s action effect
f strength
p fractiles
s standard deviation
v variation coefficient
X
x }mean value of x
mean
0 safety-factor
E strain
fJ. E strain-variation
G stress

Indices: c concrete
cr creep
k characteristic
1 lower characteristic value
o zero (eg time = 0)
p prestressing, prestressing steel
rel relaxation
s steel
sh shrinkage
t time
u upper characteristic value
x abscissa (length of the element)
y yield point
F force
M material
Q due to load
J..L friction
434 Woelfel

Table 1: Safety factors 'l F for actions

s L s u L s
acting un- favorable un- favorable
favorable favorable

self weight 1.0 1.0 1.35 1.0

(Gmean) (1.04) (0. 96) (1.33) (0.89)

prestressing 1.1 p 0.9 p 1.2 p 0.9 p

mean mean mean mean
(1.0 pk ) (1.0 Pk,l) (1.0 pk ) (1.0 Pk,l)
,s ,s

live load 1.0 - 1.5 -

Table 2: Material-safety factors '6 M

concrete reinforcing prestressing

steel steel

'6M 1.5 1.15 1.15

 Internal tendons External tendons !

A Ep s 1) A Ep s 1)

Friction
~p.~ • {1-exp[-~(atkx)]} Epo,max 1.02·10-3 0.35·Afp,~ 0.41·10-3 0.20·Afp,~

-
I
{0.2 roos
with ~ a • 35" k- 0
0.1 I
i

£po,max • 0.9·fpo.1~Es • 6.8·10-3; fpo. 1k • 1600 N/mm 2

I
I
Creep
Afp,cr • ~,cr 0.70·10-3 0.20·AEp,cr o.6o-Io- 3 0.20·AEp,cr

Shrinkage
AEp,s • i';;,s 0.30·10-3 0.35·Afp,s 0.30·10-3 0.35·Afp,s

~
Relaxation
fEp,r - 0.04 £p,x • 0.04(£po,max - ~~) 0.23·10-3 0.30·A£p,r 0.26·10-3 0.30·Afp,r lfl

fi
2.25·10-3 0.1B·Afp 2) 1. 57 .}0"3 O.l25·Afp 2) lfl
lfl

--- - ---
s·
(1q

Table 3: Losses of prestrain for internal and external tendons.

~
1) s =Standard deviation ace. [2], [3] 2) s = ys 12ts 22t .• sn2 C;.'l
CJ1
436 Woelfel

fyk ----------------------
fyk 'Ys
ULS 6pku
{
Gpkl
Gpku
SLS
{
Gpkl

prestressing x =0
t =0

Ql
1/)

5l
....
u
c:
c:
0
....
Iii due to deformation
of the section regarded

Fig 1 : Strains e: and stresses 6 in tendons

 SP 120-22

Reliability Analysis of Externally

Prestressed Concrete Bridge Girders
by A.S. Nowak, A.E. Naaman, and S.-C. Ting

Synopsis: Evaluation of existing bridges is an important part of the strat-

egy of dealing with the deteriorating infrastructure. Load and resistance
parameters are random variables because of uncertainties in load compo-
nents, material properties and dimensions. Therefore, the reliability of a
structure is a convenient measure of its performance. The load and resis-
tance models are first summarized. The major load components in bridges
are dead load and live load. The live load model is based on the weigh-
in-motion studies. Girder distribution factors were derived using special
computer procedures for bridge analysis. Behavior of composite girders is
considered using a nonlinear model. The basic characteristic of the section
is the moment-curvature relationship. The reliability is measured in terms
of the reliability index. The approach is demonstrated on evaluation of
a prestressed concrete girder bridge. Three cases are considered: original
design condition, damaged with corroded strands and repaired by exter-
nal prestressing. The load components and load carrying capacities were
evaluated and then the reliability indices were calculated for the three
cases.

Keywords: bridges (structures); damage; deterioration;

evaluation; girders; loads (forces) ; prestressed concrete;
prestressirg; reliability; repairs; serviceability;
st.remthenirg; structural analysis; unborrled prestressirg

437
438 Nowak, Naaman, and Ting

A.S. Nowak is an Associate Professor of Civil Engineering at the Univer-

sity of Michigan. He has been actively involved in the research related to
bridge design and evaluation criteria, structural reliability and LRFD for
bridges. He is a member of ACI Committees 343, 348 and 358.
A. E. Naaman is a Professor of Civil Engineering at the University of
Michigan. His major areas of expertise include partially prestressed con-
crete, external prestressing and high performance fiber reinforced cement
composites. He is an ACI Fellow and a member of committees 363, 446,
544, 549, 343 and 423.
S.-C. Ting is a doctoral candidate at the University of Michigan. He has
an extensive experience in the design of concrete structures. The topic
of his doctoral dissertation is evaluation of the effect of corrosion on the
reliability of concrete girder bridges.

INTRODUCTION

Evaluation of existing bridges is an important part of the strategy to

improve the U.S. infrastructure. There are nearly 600,000 highway bridges
in the United States. About 40 percent are classified, according to the
Federal Highway Administration's (FHWA) criteria, as deficient and in
need of rehabilitation or replacement. More than 100,000 of these are
judged to be structurally deficient because of deterioration or distress,
and another 100,000 are considered functionally obsolete or inadequate for
current requirements. The major factors which have contributed to the
present situation are: the age, inadequate maintenance, increasing load
spectra and environmental contamination. FHWA currently estimates the
cost of a bridge replacement and rehabilitation program at about 50 billion
dollars.
The disposition of the deficient bridges involves decisions ranging from
closing to traffic for replacement or repair, through posting, to unrestricted
use. Each decision involves clear economical and safety implications.
There is a need for efficient evaluation methods as well as repair and
strengthening techniques. However, bridge loads and resistance parame-
ters (dimensions, material properties) are random variables and involve a
considerable degree of uncertainty. Therefore, the structural reliability is
a convenient measure of structural performance. The reliability methods
can be used to calculate the probability of failure on the basis of the avail-
able statistical data. On the other hand, external prestressing may be
considered as a viable alternative to repair and/or strengthen prestressed
concrete girder bridges.
The analysis of a bridge involves the modeling of loads and structural
response. The parameters include axle weights and truck configurations,
truck position on the bridge, stiffness of structural members, load sharing
or interaction of members in resisting the loads, redundancies, rates of de-
 Prestressing 439

terioration, and others. The recent developments in the area of structural

analysis, material behavior, bridge load modeling and probabilistic meth-
ods may serve as a basis for rational criteria to determine the load carry-
ing capacity, reliability and remaining lifetime of the structure. There are
powerful methods available to determine the distribution of forces in the
structure, such as the finite element method (FEM) with many variants.
Significant improvements have been reported in bridge load model-
ing. Extensive live load data is available, as a result of truck surveys and
weigh-in-motion (WIM) measurements (1). There are also new results
of material and component tests. Statistical models were developed to
describe the bridge behavior under truck loading (2, 3). Probability the-
ory has developed in structural engineering especially for reliability-based
design codes (4).
Current bridge evaluation methods used in the United States are based
on deterministic approach. The American Association of State Highway
and Transportation Officials (AASHTO) Manual for Maintenance Inspec-
tion of Bridges (5) provides the procedures for most bridge inspection
and evaluation. As in the AASHTO Standard Specifications for Highway
Bridges (6), the adequacy of the structure is determined by comparing
design stresses with allowable values. Although the distribution of load
to components is the same for existing and newly designed structures,
increased allowable stresses are sometimes permitted for existing bridges.
The objective of this paper is to review the available reliability meth-
ods, and load and resistance models for bridges and illustrate their ap-
plication to a prestressed bridge strengthened by external prestressing.
The major parameters which determine the performance of prestressed
concrete girders are identified. A procedure is developed for calculation
of t~1e reliability of bridge girders in various conditions of deterioration or
damage and after their repair by external prestressing.

STRUCTURAL RELIABILITY

All quantities (except physical and mathematical constants) that enter

into engineering calculations involve uncertainties due to randomness of
design parameters (e.g. loads, material properties), imperfect models or
human errors. Because of these uncertainties, absolute reliability is not an
attainable goal. Selection of the optimum reliability level is an econom-
ical problem. Lower reliability results in frequent failures, while higher
reliability requires higher initial costs (materials and labor). Therefore,
structural reliability is a convenient acceptability criterion in the develop-
ment of design and evaluation criteria.
Classical reliability theory was developed during the 1940's for mili-
tary applications and has evolved as a result of increasing need in the area
of electronic, electrical, mechanical, and aerospace engineering. The the-
440 Nowak, Naaman, and Ting

ory is available for predicting the expected life, expected failure rate, or
expected time between breakdown of elements (systems), given some test
or failure data for the system and/or its components. In structural engi-
neering the probabilistic analysis have been developed mostly in the last
20 years. It has been applied to safety evaluation of structural members
in buildings and bridges. Little has been done, however, for the analysis
of whole structural systems. The available methodology is presented in
several recent publications, e.g. (4 ).
A structure fails when it cannot perform its function any longer, oth-
erwise it is safe. Limit states are the boundaries between safety and
failure. In bridge structures failure is defined as inability to carry traffic.
Bridges can fail in many ways, or modes of failure, by cracking, corrosion,
excessive deformations (non catastrophic failure), by exceeding carrying
capacity for shear or bending moment (collapse, catastrophic failure), by
local or overall buckling, and so on. Some members fail in brittle manner,
some are more ductile. In the traditional approach each mode of failure
is considered separately. Safety is provided by specifying the design ca-
pacity much larger than the expected loads. Probabilistic methods allow
for quantification of the safety reserve.
There are two types of limit states. Ultimate limit states (ULS)
are mostly related to the bending capacity, shear capacity and stabil-
ity. Serviceability limit states (SLS) are related to gradual deterioration,
user's comfort or maintenance costs. The serviceability limit states such
as cracking, fatigue, deflection or vibration, often govern the bridge de-
sign. The main concern is accumulation of damage caused by repeated
applications of load (trucks). Therefore, the model must include the load
magnitude and frequency of occurrence, rather than just load magnitude
as is the case in the ultimate limit states (3).
A traditional notion of the safety limit is associated with the ultimate
limit states. For example, a beam fails if the moment due to loads exceeds
the moment carrying capacity. Let R represent the resistance (moment
carrying capacity) and Q represent the load effect (total moment applied
to the considered beam). Then the corresponding limit state function, g,
can be written,
g=R-Q (1)
If g > 0 the structure is safe, otherwise it fails.
The limit state function and the associated probability of failure are
clearly defined in case of the ultimate limit states.
In general, due to uncertainties in material properties, dimensions,
workmanship, truck weights and occurrence rate, both R and Q are ran-
dom variables. Therefore, the probability of failure, Pp, is equal to,

Pp = Prob(R- Q ::=; 0) = Prob(g ::=; 0) (2)

 Prestressing 441

Let the probability density function (PDF) of R be fR and PDF of Q be

fq. Then, R- Q is also a random variable and it represents the safety
margin. Typical PDF's for R, Q and R- Q are shown in Fig. 1. The
probability of failure is equal to the shaded area.
Direct calculation of PF requires integration of convolution functions
which is usually very difficult, if not impossible. Therefore, it is conve-
nient to measure structural safety in terms of a reliability index, /3. The
reliability index is defined as a function of PF,
(3)
where q,-l = inverse standard normal (Gaussian) distribution function.
For example, f3 = 3 corresponds to PF = 0.0015, f3 = 4 corresponds to
PF = 0.000 03 and f3 = 5 corresponds to PF = 0.000 000 3.
There are various procedures available for calculation of f3. These pro-
cedures vary with regard to accuracy, required input data and computing
costs. The simplest case involves a linear limit state function (Eq. 1). If
both R and Q are independent (in the statistical sense), normal (Gaus-
sian) random variables, then the reliability index is,

f3 = 1-LR- 1-LQ (4)

uh + u~
where 1-LR = mean of R, 1-LQ =mean of Q, uR =standard deviation of R
and u Q = standard deviation of Q.
If the parameters R and Q are not both normal random variables then
Eq. 4 gives only an approximate value of f3. In such general cases, the reli-
ability index can be calculated using more advanced procedures, sampling
techniques or by Monte Carlo simulations (7). For example, an iterative
procedure is available, based on normal approximations to non-normal
distributions at the so called design point. Sampling techniques allow
for a reduction of computational effort. The basic statistical parameters
(means and standard deviations) are evaluated from a limited number of
simulations.

BRIDGE LOADS

The major load components of highway bridges are dead load, live load
with impact, environmental loads, earth pressure, and abnormal loads.
Each load group includes several subcomponents. The statistical models
for the various bridge loads are based on available data and on results
from special studies.
Further studies are required to establish the site-specific live load
model, distribution of truck weight, multiple presence in one lane and
in adjacent lanes, and transverse position of truck on the bridge. An ad-
ditional area in need of further research is dynamic load. Further field
442 Nowak, Naaman, and Ting

measurements are required to determine the effect of structural type and

surface condition. Additional studies must be carried out in order to
estimate the growth in future live loads.
Dead load, D, is the gravity load due to the self weight of the structural
and nonstructural elements permanently connected to the bridge. Because
of different degrees of variation, it is convenient to consider separately the
weight of factory made elements (steel, precast concrete members) and
weight of cast-in-place concrete members. It is assumed that D is normally
distributed. The bias factor (mean-to-nominal ratio) varies from 1.03
(factory-made members) to 1.05 (cast-in-place members), and coefficient
of variation varies from 0.04 to 0.08 (2, 7).
Live load covers a range of forces produced by vehicles moving on
the bridge and it includes a static part, L, and dynamic part, I. The
effect of live load depends on wheel force, wheel geometry (configuration),
position of the vehicle on the bridge (transverse and longitudinal), number
of vehicles on the bridge (multiple presence), stiffness of the deck (slab),
and stiffness of the girders.
Ghosn and Moses (1) developed a live load model to predict the maxi-
mum 50 year load on a bridge, in terms of bending moment in the girders.
The model is based on weigh-in-motion (WIM) data. By the use of strain
measuring instrumentation, existing bridges were used as equivalent static
scales to measure vehicle weights. Tape switches and timers were included
to determine the vehicle speed, axle spacing, and arrival time in each lane.
Because the system was inconspicuous, truck drivers did not avoid the
"weigh station" and hence the sample can be considered unbiased.
The major factors affecting the dynamic load on a bridge include sur-
face condition (bumps, potholes), natural frequency of the bridge (span
length, stiffness, mass), and dynamics of the vehicle (suspension, shock
absorbers). It is practically impossible to predict the percentage of dy-
namic load contributed by each of these three factors. In addition, the
dynamic load effect may be changed by superposition of several trucks
(multiple presence) or even by several axles of the same truck.
The bias factors (mean-to-nominal ratios) for live load vary depending
on span length. Coefficients of variation vary from 0.2 to 0.3.

RESISTANCE MODELS

The capacity of a bridge depends on the resistance of its components and

connections. The component resistance is determined mostly by material
strength and dimensions. The variation of resistance has been modeled
by tests, observations of existing structures and by engineering judgment.
The information is available for the basic structural materials and com-
ponents. However, bridge members are often made of several materials
(composite members) which require special methods of analysis. Verifica-
 Prestressing 443

tion of the analytical model may be very expensive because of the large
size of bridge members. Therefore, the available data about resistance is
still incomplete.
Typical stress-strain curves for concrete, reinforcing steel and pre-
stressing steel are shown in Figs. 2, 3 and 4, respectively.
The ultimate flexural capacity of prestressed concrete bridge girders
can be derived using a nonlinear model of performance. A computer pro-
gram has been developed (7). The statistical data is taken from (8) and
(9). Two phases of the performance are considered: uncracked section and
cracked section. The strains are assumed to be linearly distributed. The
section is divided into a number of horizontal strips of small depth. For
given strains, stresses are calculated using material stress-strain curves.
The bending moment is calculated as the resultant of the internal stresses.
The moment-curvature is obtained by incremental increases in load (bend-
ing moment).
The moment changes under cyclic loading (trucks). If the total bend-
ing moment, Mq, exceeds the cracking moment, Mer, the section cracks
and the tensile strength of concrete is reduced to zero. The crack stays
open any time Mq exceeds the decompression moment, Md (if Mq < Md,
then all concrete is compressed, if Mq > Md then the crack opens).
Moment-curvature relationships have been derived for typical
AASHTO sections, strengths of concrete and levels of prestress. The
results show that the mean-to-nominal ratio of the ultimate moment is
1.04 and the coefficient of variation is about 0.035. The coefficient of
variation is very small because all sections are under-reinforced and the
ultimate moment is controlled by the prestressing tendons.

EXAMPLE OF RELIABILITY ANALYSIS

OF A DAMAGED GIRDER

The reliability analysis is demonstrated on a typical prestressed concrete

composite bridge girder with a span of 80 ft (24.4 m). The bridge cross
section is shown in Fig. 5. Each girder is a standard AASHTO Type IV
girder with a cast-in-place reinforced concrete slab. The slab thickness
is 8 in (20.3 em). The bridge was unshored during construction. Both
the precast beam and the slab are made out of normal weight concrete
with unit weight of 150 pcf (23.6 kN/m 3 ). An asphalt topping was added
on top of the slab leading to an additional dead load of 250 plf (3.65
kN/m). The following properties were assumed in the initial design: con-
crete compressive strength of the precast girder f~ = 5 000 psi (35 MPa);
compressive strength of the cast-in-place slab= 4 000 psi (28 MPa). The
design followed the AASHTO specifications with an HS20 truck live load-
ing. It led to a prestressing force at midspan consisting of 28 strands at an
eccentricity of 20.16 in (51.2 em) from the centroid of the precast section.
444 Nowak, Naaman, and Ting

The strands were 1/2 inch (12.7 mm) in diameter and had a specified ul-
timate strength of 270 ksi (1862 MPa). The computed effective stress in
the steel after losses was 144 ksi (993 MPa). Details of the strand layout
for the midspan and the support sections are given in Fig. 6. The steel
profile is selected to have two draping points each at 30ft (9.14 m) from
the support. Twelve strands are draped and sixteen strands are straight.
Various limit states and conditions were considered: ultimate moment,
cracking, maximal crack width, fatigue in concrete, fatigue in steel and
live load deflection. Reliability indices were calculated using Eq. 4, with
definition of R given in Table 1, and Q defined as applied moment (dead
load, live load and impact).
The moment-curvature relationship for the composite girder was de-
rived using the available statistical data on dimensions and material prop-
erties as reported in Table 2. The resulting average curve is shown in Fig. 7
as a solid line while the dashed lines correspond to the average plus or mi-
nus one standard deviation. The girder midspan moments were calculated
for dead load, Mv, superimposed dead load, and live load plus impact
ML+l· Observed nominal and mean values with bias, standard deviation,
and coefficient of variation are reported in Table 2. Corresponding values
of the reliability index (3 for ultimate strength and various serviceability
limit states are reported in Table 3. Note that the probability of failure or
survival with respect to one limit state can be estimated from the values
of the reliability index. Typical examples are given in Table 4. It can
be observed for instance that, at time of completion, the probability of
girder failure by fatigue of concrete under normal loading conditions is
less than three in ten millions. It is noted that the nominal moment car-
rying capacity of the girder is 5180 k-ft (7 023 kN-m) and the safety factor
representing the ratio of nominal resistance in bending to the maximum
service moment was calculated as 2.00.
In this example it is assumed that due to some severe environmental
exposure to corrosion, 8 straight strands from the layer of reinforcement
closest to the bottom fiber were damaged, and snapped off. The analysis
of the bridge in this damaged condition indicated that: 1) the dead load
moment exceeded slightly the cracking moment, 2) the girder would crack
under live load, 3) the cracks will open at each application of the loading,
4) the stress range in the prestressing strands due to the full application
of live loads is high enough to raise concern about fatigue.
The safety factor given by the ratio of the nominal moment resistance
of the damaged girder to the maximum service moment is 1.59. Clearly
however, due to the cracking condition and the possible fatigue in the rein-
forcement the bridge needed repair. Application of the reliability analysis
to the damaged girder led to the moment curvature relationship shown in
Fig. 8. The analysis also led to a mean resistance of 3830 k-ft (5193 kN-
m) and a corresponding reliability index for the ultimate strength limit
 Prestressing 445

state f3 =6.94 (see Table 3). This is equivalent to a minute probability of

failure (collapse) (Table 4). However, analysis of other limit states indi-
cated a sure probability of cracking and a value of reliability index of 0.16
for fatigue indicating a probability of failure by fatigue close to 50% at
two million cycles of full live load. Variations of stress increments in the
concrete and the prestressing steel due to the application of live load are
shown if Fig. 9 and 10. It can be observed that stress increments in the
prestressing steel of the damaged girder can be significant and should be
accounted for in the evaluation of repair strategies.
Although a conventional analysis of the damaged girder may lead to
the decision to conduct emergency repair and close the bridge during re-
pair, the reliability analysis allows the bridge engineer to make a more
rational decision. In this example, since the reliability index (for ulti-
mate) of the damaged structure is still larger than 3.5 (the minimum
index required for new members), the bridge does not need to be closed,
and repair can be made at a more reasonable time and cost.
Repair consisted of removing the spalled concrete cover, cleaning the
damaged area, patching with a latex modified mortar to initial shape,
and replacing the damaged strands by extemal prestressing. Extemal
prestressing was achieved using two Dywidag prestressing bars having a
nominal area of 1.485 in 2 (9.58 cm 2 ) each, and a nominal strength of 150
ksi (1 034 MPa). Figures 11 and 12 illustrate the method of anchoring
the bars to cast-in-place reinforced concrete blocks attached to the lower
portion of the precast beam. The sudace of the beam over which the
blocks were poured was roughened, and dowel bars were run through the
web to improve horizontal shear transfer. The anchor blocks were placed
at 15ft (4.57 m) from each support to allow the dowel reinforcement to
be placed through drilled holes in the web of the beam located between
the draped and the straight tendons (Fig. 12).
The repaired girder provided the same cracking strength under ser-
vice load as the original girder. Existing cracks which formed while the
bridge was damaged will open slightly during application of full live load.
However the crack widths will be very small and the stress range in the
reinforcement will be much smaller than the limit allowable for fatigue
endurance. The moment curvature relationship of the repaired girder is
shown in Fig. 8 and compared to that of the original and the damaged
girder. Its nominal resistance is 5 045 k-ft (6 840 kN-m); values of the reli-
ability index for the various limit states are given in Table 3. These values
are comparable to those obtained for the original design and suggest that
the repair procedure is acceptable.

CONCLUDING REMARKS

The reliability analysis of a structure being new, old, or damaged, for var-
ious strength and serviceability limit states provides not only convenient
446 Nowak, Naaman, and Ting

measures of its performance (through the reliability index) but is also a

rational and efficient procedure for its evaluation. When applied to the use
of existing bridges, such analysis should allow bridge engineers to make
rational decisions ranging from unrestricted use, to closing to traffic for
replacement or repair. Once the methodology has been perfected at the
research level and calibrated to reflect real life situations, its wide scale
use in practice should be encouraged.

ACKNOWLEDGEMENT

The presented research was partially supported by the National Science

Foundation under grant ECE-8413274 with Program Director John B.
Scalzi and grant MSM-8613440 with Program Director Dr. G. Albright.
Their support is gratefully acknowledged.

REFERENCES

1 Ghosn, M. and Moses, F., 1984, "Bridge Load Modeling and Reliabil-
ity Analysis," Report No. R 84-1, Department of Civil Engineering,
Case Western Reserve University.
2 Nowak, A. S. and Zhou, J. H., 1985, "Reliability Models for Bridge
Analysis," Report No. UMCE85-9, University of Michigan, March.
3 Nowak, A. S. and Grouni, H. N., 1988, "Serviceability Considerations
for Guideways and Bridges," Canadian Journal of Civil Engineering,
Vol. 15, No. 4, August, pp. 534-538.
4 Thoft-Christensen, P. and Baker, M. J., 1982, Structural Reliability
Theory and It3 Applications, Springer-Verlag, p. 267.
5 AASHTO, 1983, "Manual for Maintenance Inspection of Bridges,"
American Association of State Highway and Transportation Officials,
Washington, D. C.
6 AASHTO, 1983, "Standard Specifications for Highway Bridges,"
American Association of State Highway and Transportation Officials,
Washington, D. C., p. 394.
7 Nowak, A. S. et al., 1988, "Risk Analysis for Evaluation of Bridges,"
Report UMCE 88-7, Department of Civil Engineering, University of
Michigan, Ann Arbor, November.
8 Ellingwood B. et al., 1980, "Development of a Probability Based Load
Criterion for American National Standard A58," National Bureau of
Standards, NBS Special Publication 577, Washington.
9 Siriaksorn, A. and Naaman, A. E., 1980, "Reliability of Partially Pre-
stressed Beams at Serviceability Limit States," Report No. 80-1, Uni-
versity of Illinois at Chicago Circle, Chicago, June.
 Prestressing 44 7

Table 1 - Definition of R in the Reliability Analysis

limit state R

1. ultimate moment carrying capacity

2. cracking cracking moment
3.max.crack width moment causing critical crack width (0.016")

4. fatigue in concrete moment causing critical stress range (0.4J:-

fmin/2, where fmin =min. stress in concrete)

5. fatigue in steel moment causing critical stress range (0.1fpu)

6. max. live load moment causing critical deflection (L/180)

deflection

Table 2 - Statistical Data for Basic Variables

item description nominal bias mean v (1

1 girder dead load, kip-ft 657.5 1.030 677.2 .040 27.1

2 additional dead load, kip-ft 789.7 1.037 819.2 .083 67.6
3 live load plus impact, kip-ft 889.0 0.819 728.0 .240 174.7
4 internal strand strength, ksi 270 1.041 281 .025 7.0
5 external tendon strength, ksi 150 1.087 163 .042 6.8
6 girder concrete f~ = 5 000 psi 4250 .948 4028 .150 604
7 slab concrete J:
= 4 000 psi 3400 .997 3390 .180 610

8 in. strand effective depth, inch 49.43 .996 49.25 .011 0.52
9 ex. tendon effective depth, inch 37.00 .995 36.82 .014 0.52
10 slab thickness, inch 8.00 1.004 8.03 .059 0.47
448 Nowak, Naaman, and Ting

Table 3 - Reliability Index j3 for Various

Limit States and Conditions

item limit state original damaged repaired

1 ultimate 12.10 6.94 11.65

2 cracking 1.62 -1.71 1.76
3 maximal crack width 3.53 -1.34 3.05
4 fatigue in concrete 5.02 2.47 4.80
5 fatigue in steel 4.29 0.16 4.30
6 maximal live load deflection 9.04 4.19 8.92

Table 4 - Reliability Index j3 versus Reliability S and

Probability of Failure Pp

Reliability Reliability S ( = 1 - Pp) Probability

Index j3 or Probability of Safety of Failure Pp

0.0 0.500 0.500x1o+o

0.5 0.691 0.309x1o+o
1.0 0.841 0.159x1o+o
1.5 0.933 2 0.668x1o- 1
2.0 0.9772 0.228x1o- 1
2.5 0.993 79 0.621x10- 2
3.0 0.99865 0.135x1o- 2
3.5 0.999 767 0.233x10- 3
4.0 0.999968 3 0.317x1o- 4
4.5 0.999996 60 0.340x1o-s
5.0 0.999 999 713 0.287x1o-s
5.5 0.999 999 981 0 0.190x10- 7
6.0 0.999 999 999 013 0.987x1o- 9
6.5 0.999 999 999 959 8 0.402x10- 10
7.0 0.999 999 999 998 72 0.128x1o-n
7.5 0.999 999 999 999 9681 0.319x10- 13
8.0 0.999 999 999 999 999 389 0.611 X 10- 15
 Prestressing 449

Load and Resistance

Fig. 1 - Probability Density Functions for R, Q and R - Q

1.0 2.0 3.0

Strain, €c

Fig. 2 - Stress-Strain Curves for Concrete

450 Nowak, Naaman, and Ting

100

80

60
]
vr
"'
~
""
20

0
0.00 0.01 0.02 0.03 0.04 0.05
Strain

Fig. 3 - Stress-Strain Curves for Reinforcing Steel

300

prestressing strand

200
]
vi
"' prestressing bar
~
100

0.02 0.03 0.04 0.05 0.06

Strain

Fig. 4 - Stress-Strain Curves for Prestressing Steel

 Prestressing 451

L~--------------32'-----------------~
[r3'--+•·-----12'-----+-----12'-------t-3' ,.

Fig. 5 - Cross Section of the Bridge Considered

in the Example

Midspan Section End Section

1-------6'9''----------1 1 - - - - - - 6'9"---------11

Fig. 6 - Cross Section of the Girder with Layout of

Prestressing Tendons
452 Nowak, Naaman, and Ting

6.------------------------------------------,
__________
,-.-_:-::.:-::.;.-_-
~~ ,..-
'1'/
-----------

Mean Value

~82!:!.Y.aJ!:!.e.:!:.g>:
~2!:!.Y.aJ!:!.~g>:

100 200 300 400 500 600 700 800

Curvature rp, 10-a/InCh

Fig. 7 -- Moment-Curvature Relationship for the Girder

6,------------------------------------------,

a Original
181Damaged
A Repaired

o~~~~~~~~~~~~~~~~~~~~~~
0 100 200 300 400 500 600 700 800
Curvature rp, 10-a/inch

Fig. 8 -- Moment-Curvature Curves for the Original,

Damaged and Repaired Girder
 Prestressing 453

0.8

1 M 0 -orlglnal (slab)
0.7
2 t.f. -original {girder)
0.6
3 M.-damaged {slab)
]
~
0.5 4 M • -damaged {girder)

j
Ill
0..4
5 t.f•-repaired {slab)
6 M. -repaired (girder)
Ill 0.3

~ 0.2

0.1

0.0
0 0.2 0..4 0.6 0.8
Bending Moment, (M - M0 )/ML+t

Fig. 9 - Typical Stress Increment in Concrete Top Fiber

with Live Load Moment

30

25
1 M,.. -original
] 2 t.f,..-damaged
20
"EG> 3 M,..-repaired {In)

j Ill
15
4 t.f,..-repaired (ex)

Ill

~
10

0.2 0..4 0.6 0.8

Bending Moment, (M - M0 )/ML+t

Fig. 10 - Typical Stress Increment in Prestressing Steel

with Live Load Moment
454 Nowak, Naaman, and Ting

---6'9"-------l•l
~-----,-
---r-
8"

Anchor
Blocks and
External
Tendons

54"

6@2" = 12"

Fig. 11 - Cross Section of the Repaired Girder

RC t= 8"

Holding Down Point

1------15'------~·'
Anchor Block for
for Internal Strands External Tendons

Fig. 12 - External Tendon Arrangement

in the Repaired Girder
 SI (metric) Tables 455

CONVERSION FACTORs-INCH-POUND TO Sl (METRIC)*

To convert from to multiply by

Length

inch ............................ millimeter (mm) ..................... . 25.4Et

foot ............................... meter (m) ........................ . 0.3048E
yard .............................. meter (m) ........................ . 0.9144E
mile (statute) .................... kilometer (km) ...................... . 1.609

Area
square inch ................... square centimeter (cm 2) • • • • • • • • • • • • • • • • • • • 6.451
square foot ..................... square meter (m') .................... . 0.0929
square yard ..................... square meter (m 2 ) • • • • • • • • • • • • • • • • • • • • • 0.8361

Volume (capacity)

ounce ........................ cubic centimeter (em") .................. . 29.57

gallon .......................... cubic meter (m'):j: .................... . 0.003785
cubic inch . . . . . . . . . . . . . . . . . . . cubic centimeter (em") .................. . 16.4
cubic foot ....................... cubic meter (m") ..................... . 0.02832
cubic yard ....................... cubic meter (m"):j: .................... . 0.7646

Force
kilogram-force ..................... newton (N)........................ 9.807
kip-force .......................... newton (N) ........................ 4448
pound-force ....................... newton (N)........................ 4.448

Pressure or stress (force per area)

kilogram-force/square meter ......... pascal (Pa) ....................... . 9.807

kip-force/square inch (ksi) ........ megapascal (MPa) .................... . 6.895
newton/square meter (N/m 2 ) • • • • • • • • • pascal (Pal ....................... . 1.000E
pound-force/square foot . . . . . . . . . . . . pascal <Pal ....................... . 47.88
pound-force/square inch (psi) ...... kilopascal (kPaJ ..................... . 6.895
456 SI (metric) Tables

To convert from to multiply by

Bending moment or torque

inch-pound-force ............... newton-meter (Nm) ................... . 0.1130

foot-pound-force ............... newton-meter (Nm) ................... . 1.356
meter-kilogram-force ........... newton-meter (Nm) ................... . 9.807

Mass
ounce-mass (avoirdupois) ............. gram (g) ........................ . 28.34
pound-mass (avoirdupois) ........... kilogram (kg) ............. ·......... . 0.4536
ton (metric) ....................... megagram (Mg) ..................... . I.OOOE
ton (short. 2000 Ibm) .............. megagram (Mg) ..................... . 0.9072

Mass per volume

pound-mass/cubic foot ...... kilogram/cubic meter (kg/m'')................ 16.02
pound-mass/cubic yard ...... kilogram/cubic meter (kg!m·') . . . . . . . . . . . . . . . . 0.5933
pound-mass/gallon .......... kilogram/cubic meter (kg/m")................ 119.8

Temperature§
deg Fahrenheit (F) ................. deg Celsius (C) ................ tc = (IF - 32), 1.8
deg Celsius (C) .................. deg Fahrenheit (F) ................. tF = 1.8tc "'" 32

*This selected list gives practical conversion factors of units found in concrete tech-
nology. The reference source for information on SJ units and more exact conversion
factors is "Standard for Metric Practice.. ASTM E 380. Symbols of metric units are
given in parentheses.
+E Indicates that the factor given is exact.
t One liter (cubic decimeter) equals 0.001 m" or 1000 em".
~ These equations convert one temperature reading to another and include the neces-
sary scale corrections. To conver: a difference in temperature from Fahrenheit degrees
to Celsius degrees. divide by 1.8 only. i.e .. a change from 70 to 88 F represents a change
of 18 For 18/1.8 = 10 C deg.
 457

INDEX

Anchorage (structural), 61 Fatigue (materials), 107

Fenves, G.L., 389
Bakeer, R.M., 409 Finite element method, 389
Beams (supports), 107, 339 Flexural strength, 207, 339
Beaupre, R.J., 257 Friction, 83
Bending moments, 355
Box beams, 1, 61, 159, 175, 185, Galvanized materials, 83
229, 257, 315, 355, 389 Gauthier, Y., 355
Breen, J.E., 257, 315 Gauvreau, P., 289
Bridges (structures), 1, 61, 83, Girders, 185, 305, 437
125, 145, 159, 175, 207, 229, Grout, 83
257. 289, 305, 315, 355, 375, Grouting, 125, 175
389, 437
Bruggeling, A.S.G., 61 Hangars, 185
Building codes, 425 High-strength concretes, 145
Burgoyne, C.J., 107 History, 1, 61, 125

Cable-supported structures, 1, Ivkovic, M., 185

61, 159
Cantilever beams, 159 Jacking, 409
Cantilever bridges, 1, 145, 175 Jartoux, P., 83
Causse, G., 175 Joints (junctions), 175
Coatings, 229
Composite construction (concrete Klaiber, F.W., 207
and steel), 1, 289 Kreger, M.E., 257, 315, 389
Computer programs, 355, 409
Concrete construction, 1, 125, Lacroix, R., 83
145, 159 Lateral pressure, 229
Concrete slabs, 289 Limit design method, 125
Corrosion, 125 Limit state design, 425
Corrosion resistance, 107 Loads (forces), 207, 229, 437
Cracking (fracturing), 339 Load tests (structural), 229
Creep properties, 107, 185 Long span, 83, 185

Damage, 437 MacGregor, R.J.G., 315

Das, S.C., 409 Matejowsky, A., 159
Deterioration, 229, 437 Menn, C., 289
Domes (structural forms), 409 Models, 257, 289, 315
Ductility, 315 Modulus of elasticity, 107
Ducts, 83 Moment-curvature relationship, 355
Dunker, K.F., 207 Muller, J., 355
Durability, 107, 289
Naaman, A.E., 339, 437
Economics, 125 Nagele, J.M., 409
Eibl, J., 375 Nowak, A.S., 437
El-Habr, K.C., 389 Nuclear reactor containment, 409
Epoxy resins, 159, 229
Evaluation, 409, 437 Osborn, A.E.N., 229
458 Index

Partial prestressing, 339 Shells (structural forms), 409

Perisic, Z. , 185 Shotcrete, 229
Plastics, polymers, and resins, Sowlat, K., 305
Post-tensioning, 125, 145, 159, Spans, 207
175, 207, 229, 257, 305, 315, Stadiums, 145
355, 389 Strains, 125
Powell, L.C., 257 Strength, 107, 315
Precast concrete, 145, 159, 175, Strengthening, 207, 437
185, 257, 305, 315, 355, 389 Stresses, 207
Preston, H.K., 229 Stress relaxation, 107
Prestressed concrete, 1, 61, 83, Stress-strain relationships,
107, 125, 145, 185, 289, 305, 355, 425
339, 375, 409, 425, 437 Structural analysis, 339, 389,
Prestressing, 1, 61, 83, 107, 409, 437
125, 145, 159, 175, 185, 339, Structural design, 1, 125, 145,
375, 389, 409, 425, 437 175, 185, 207, 257, 305, 315,
Prestressing steels, 1, 61, 83, 339, 355, 375, 425
145, 159, 175, 229, 257, 305, Supports, 61
315, 355, 375, 389, 409 Svensson, H.S., 125

Rabbat, B.G., 305

Reliability, 437 Tests, 207, 257, 289, 305,
Renovating, 207 315, 375
Repairs, 229, 437 Thao, P.X., 145
Research, 207, 305 Thermal properties, 107
Reviews, 1, 61, 83 Ting, S.-C., 437
Roofs, 185
Unbonded prestressing, 1, 61, 83,
Safety factor, 425 125, 145, 159, 175, 185, 207,
Sanders Jr., W.W., 207 257, 289, 305, 315, 339, 355,
Scale (ratio), 289, 305 375, 389, 425, 437
Segmental construction, 1, 61,
145, 159, 175, 257, 305, 315, Viaducts, 145
355, 389 Virlogeux, M.P., 1
Serviceability, 61, 185, 289,
339, 355, 425, 437 Woelfel, E., 425