You are on page 1of 10

Deflections of Precast Pretensioned Beams

John Connal, Industry Director – Structures, Maunsell Australia Pty Ltd

SYNOPSIS

Many medium span bridges are built with precast pretensioned beams. The
majority of the bridges built in Australia are up to 35m span and the most
common of these is in Super T form.

The AS5100 code requires these beams to be built with no sag. However
even when the provisions of the code have been applied, despite predictions
of long term hog and adequate strength, beams of some bridges have
sagged. Why does this happen?

This paper explores, by way of actual examples, the reasons why some
bridge beams, under some conditions, do not behave as the code predicts. It
considers effects of creep and differential shrinkage, the effects of
prestressing and non-prestressed reinforcement, and the influence of
environmental conditions.

1.0 DESCRIPTION OF THE PROBLEM

There have been a number of occasions when bridges comprised of


pretensioned beams have sagged, notwithstanding the designer of these
bridges has not predicted this. The number of occasions this has occurred is
difficult to quantify because when it does happen either a dispute may erupt or
some embarrassment may be experienced, inhibiting wider discussion of the
problem.

The author knows intimately of two such occasions and two other cases in
less detail. This paper is based on two particular experiences but the projects
remain un-named in order to bring the issue to discussion in a wider forum.

1.1 NSW Bridge


This example involved pretensioned Super T beams of closed section with
spans up to 35.6m and beam spacing up to 2.45m. It was designed to the
Australian Bridge Design Code [AUSTROADS (1992)] and for T44/L44 live
loading and HLP320 heavy platform loading. The beams were 1500mm deep
and heavily pretensioned and also included 6/Y32 bars at mid span and these
were curtailed before the ends of the beams. However a construction error
omitted two of the bars, so only 4/Y32 bars were actually included in the
beams. The design conservatively predicted hogs of 75mm for these beams,
however later calculation showed that the predicted hogs should have been of
the order of 60mm, and that the drawings indicated a slightly larger hog
because the usual tendency was for designs that are in accordance with the
Australian Bridge Design Code (ABDC) to under-estimate hogs.

Austroads 2006 – Deflections of Pretensioned Beams - Connal Page 1 of 10


There was never any evidence that established the hog measurement just
after transfer, hence the sag behaviour that these beams was never fully
clear. However just prior to casting the deck, the hogs were measured at 10-
30mm, so over the time between casting the beams and casting the deck, the
beams probably sagged by 30-50mm. There was never a satisfactory
explanation provided for the sag of the beams. Their behaviour was never
predicted by the ABDC.

It was found that the behaviour of these beams could not be predicted by the
information and design methodologies contained in the ABDC. The beams
were cast in about December 1996 but the deck was not cast on them until at
least February 1997. The beams were left near the bridge site exposed to the
weather for about three months over summer.

Whilst the ABDC requires that, in order that deflections do not detract from the
appearance of a structure, sag should not occur in beams under service
conditions. In this instance it was concluded that the strength of the beams
were not compromised and notwithstanding a slight sag in the bridge, it was
deemed serviceable.

1.2 Victorian Bridge


This bridge is a three-span continuous structure with spans of 40m/32m/34m,
with a skew of about 40 degrees. The beams were 1500mm deep, with
widths that varied between 1.75m and 2.23m for outer and inner beams
respectively. They were designed for SM1600 loading to AS5100-Bridge
Design (Standards Australia 2004). In the case of this bridge the beams in the
longer first span (Span 1) sagged.

It was established that the construction sequence for the bridge was not fully
followed, and more deck slab weight was applied to the beams in their simply
supported state prior to continuity being established. However the other
possible differences from the design assumptions were investigated and
eliminated. Namely:
• Prestress missing or not stressed to the values specified.
• Concrete cross section larger than designed (eg. smaller internal void),
increasing the dead weight of the beams and reducing the effectiveness of
the prestress (P/A effect).
• Curing regime compromised.
• Exposure to temperature and humidity extremes.
• Significant differential shrinkage in the upper portions of the precast
beams relative to the lower portions.
• Abnormally high concrete density, and
• Prestress not positioned in the as-designed positions within the beam
cross section.

The beams did not hog upwards as much as expected and this lower-than-
predicted hog also meant that the cast-in-situ deck concrete was thicker than
designed. This caused an increase in downward deflection of up to 10mm.
The departure from the designed bridge construction sequence also caused
an increase in downward deflection of about 10mm. The comparison of beam

Austroads 2006 – Deflections of Pretensioned Beams - Connal Page 2 of 10


deflections is given below based on the outer beams of the worst case of
beam deflection for Span 1. In Table 1 below, the negative values represent
an upward hog and positive values represent a downward sag.

Table 1 — Beam Deflections for Span 1 — Outer Beams

Beam deflections in mm
Hog @ After deck4 After superimposed
28 day hog2
transfer cast loads5
Designed -60 -90 -55 -40
Revised design1 -60 -90 -35 -25
Actual -45 -233 32 42
Notes:
1. As-designed valves modified for revised deck construction sequence and thicker deck slab.
2. Age of beams at deck casting date assumed by the design.
3. This represents a sag of 22 mm from transfer to 28 days. Note however, that hogs were not measured at
the same reference points or with the same equipment. Also there is no correction of hog measurement
for differential temperature effects that may have prevailed at the time of measurement of the hog.
4. The deck was cast 118 days after casting the beams.
5. These are barriers, asphaltic concrete surfacing.

Notwithstanding the construction variations from the design, the basic reason
for the small sag in the beams of Span 1 is the smaller than anticipated hog in
the beams at the time of casting the deck. The beams were up to 118 days
old when the deck was cast and were stored either in the precast yard or on
site or in position on the piers and abutment over this time during summer.
Whilst the summer was not particularly hot, the beams were exposed over a
relatively long period of time to relatively severe drying conditions. Like the
previous example, the beams of this bridge were considered to have been
designed in accordance with the code and were considered to have sufficient
strength and safety. The slight sag in the beams did not compromise
clearance and were considered to have adequate serviceability,
notwithstanding the sag does not comply with the code.

It was considered there are likely to be a number of factors that influenced the
behaviour of these beams and these are investigated to seek an explanation
of the causes. These are described in the following sections.

2.0 LOAD SHEDDING TO BONDED REINFORCEMENT IN THE


SECTION DUE TO CREEP AND SHRINKAGE

This phenomenon has been recognised for a long time in the design of doubly
reinforced concrete beams using working stress design methods, and in the
creep buckling of reinforced concrete columns. It traditionally was dealt with
in an empirical way in the working stress design of doubly reinforced concrete
beams by using a modular ratio of 2n-1 when converting the compression
steel in the section to a concrete equivalent. This recognised that the
compression steel actually attracts more load than one would predict if creep
and shrinkage was ignored.

Essentially four time dependent phenomena are occurring in a pretensioned


prestressed concrete beam. The concrete will creep under the sustained load

Austroads 2006 – Deflections of Pretensioned Beams - Connal Page 3 of 10


of prestress and other permanent loads. The concrete will also undergo
shrinkage strains. The concrete matures with time and its elastic modulus
increases, and relaxation occurs in the prestressing steel. The gradual
change in strain in the concrete causes changes of stress in the bonded
prestressing strand as does its relaxation. However when bonded non-
prestressed (or passive) reinforcement is also present in the cross section, it
too is subject to changes in stress. In effect, as the concrete shortens due to
compressive creep and shrinkage, both the prestressed and non-prestressed
reinforcement is compressed. This results in a loss of prestress (in addition to
the relaxation of the prestressing strand) and a gradual build-up of
compression in the non-prestressed, passive reinforcement.

This effect is documented in Gilbert (1988), where several numerical models


are suggested to predict the behaviour of various concrete elements. These
numerical models are complex and usually employ a time history or step-by-
step analysis of the stress and strain history of the concrete element over
time. The numerical accuracy of the stress histories depends on the accuracy
of the material properties used as data. If the variability of the estimates of
material properties is high, then so too will be the variability of the predictions.
In the case of the prediction of hogs in pretensioned, prestressed and
reinforced concrete beams, the hog is the result of the algebraic sum of two
larger quantities; the downward deflection of the beam self weight and the
upward deflection due to the pretensioning force. In these circumstances the
fluctuations in these quantities will yield a large variation in the resulting hog.

The variability in the material properties of concrete also contributes to the


inaccuracy that can result in the deflection calculations. The AS5100 code
does recognise that creep and shrinkage can vary from the code predictions
by ±30%. It also suggests that elastic modulus can vary by ±20%. These are
quite large variations and suggest that accurate numerical modelling can only
be based on relevant test data for the concrete in question.

Gilbert (1988) suggests that the Age-adjusted Effective Modulus Method


(AEMM) is usually the most satisfactory modelling method for manual
calculation. The AEMM approach has been used to investigate this behaviour
for the case of the beams of Span 1 of the Victorian bridge example above.

The calculations involved with the AEMM approach are complex and the
results from this approach should be considered approximate, however they
do match the same trends of the Gilbert (1988) work, and they are considered
to be reasonable representations of the beam behaviour based on the
assumed properties for the constituent steel and concrete materials. Two
other analysis methods have also been used to provide a confirmation of the
AEMM results. The in-house Maunsell software “CREAP” and the step-by-
step method [SSM, also indicated by Gilbert (1988)], have also been used to
analyse the beam. Both these methods indicate the same trends for the
beam behaviour, to give some comfort that the predictions are reasonable.

The modelling conducted has, by necessity of the complexity of the


calculations, made some approximations. The properties selected for the

Austroads 2006 – Deflections of Pretensioned Beams - Connal Page 4 of 10


steel and for the behaviour of the concrete are essentially taken from AS5100,
but the initial modulus (Ec) of the concrete has been taken as 35 GPa (this is
a value between the Ec that AS5100 would predict for concrete with a cylinder
strength of 40 MPa, as designed and achieved at transfer, and the Ec that the
code would predict for a cylinder strength of 69 MPa; the actual average
measured value at 28 days). This may underestimate the initial elastic
deflections by about 10%. However the AEMM approach does not allow for
the change in elastic modulus with time. Therefore a consistent value, slightly
higher than the calculated value for 40 MPa concrete, was used to allow for
the increase in modulus with time. The actual beams had debonded
prestressing strands. However the effect of debonding overly complicates the
numerical modelling and was not included. As a result, it is expected that the
deflections would be slightly over-estimated. After both these approximations
are considered, it is suggested that the computational results from this
analysis could be expected to be reasonably representative, but certainly not
exact!

The results of this work are shown in Table 2, and shown in graphical form in
Figure 1. Included with these results are some sensitivity analyses together
with the deflections predicted by the original design process with a ±30%
range, the hogs shown on the drawings, and the measured hog.

Table 2 — Deflection Response with Time (mm)


Time (days)
0 30 60 90 120
3,4
Indicated on drawings 60 90 - 110 -
As-designed plus 30% 79 101 - 114 -
As-designed 61 78 - 88 -
As-designed minus 30% 43 55 - 62 -
Measured 45 23 - - 191
AEMM predictions assuming a constant average hypothetical thickness of 125 mm5
Ignoring passive reinforcement 58 74 79 82 84
With all passive reinforcement 55 66 70 73 74
(8Y12 top bars, 2Y36 bottom bars)2
With only bottom passive reinforcement 55 64 67 69 70
(2Y36 bottom bars)
With hypothetically increased passive 48 50 51 51 51
reinforcement (8Y12 top bars/ 6Y36 bottom
bars)
With hypothetically increased passive 47 47 47 47 47
reinforcement (6Y36 bottom bars)
Notes:
1. The beam deflection was measured at transfer and at about 28 days and then not measured again until
after the deck was cast (at 118 days). The value given at 120 days is estimated, based on anticipated
elastic deflection due to deck casting.
2. Actual passive reinforcement in beams was 8Y12 bars in the top flange and 2Y36 bars in the bottom of the
beam. Web reinforcement has been ignored in this analysis.
3. The hogs shown on the drawings were estimations based on the properties assumed in the design and for
both inner and outer beams. The inner and outer beams have different top flange widths and actually
deflect slightly differently. This was ignored on the drawings for simplicity.
4. The hogs shown on the drawings were rounded up to slightly higher values because past experience has
almost always shown that the beams hog more than that predicted by the code.
5. The hypothetical thickness used in the AEMM calculation past experience has almost always shown that
the beams hog more than that predicted by the code.

Austroads 2006 – Deflections of Pretensioned Beams - Connal Page 5 of 10


120 As-designed plus 30%

Indicated on drawings
100
As-designed

80 AEMM ignoring passive


Deflection (mm)

reinforcement

AEMM with all passive


reinforcement (8Y12 top bars, 2Y36
60 bottom bars)
AEMM with only bottom passive
reinforcement (2Y36 bottom bars)

40 As-designed minus 30%

AEMM with hypothetical increase


in passive reo (8Y12 top bars and
20 6Y36 bottom bars)
AEMM with hypothetical increase
in passive reinforcement (6Y36
bottom bars)
0 Measured

0 30 60 90 120
Time (days)

Figure 1 Deflection Response – “As-designed” versus “AEMM Calculated” versus


“Measured”

The AEMM approach predicts the same behaviour as that predicted by design
using AS5100. Lesser deflection is predicted but the behaviour is similar.
The AS5100 design predictions with a ±30% range encompass the range of
values predicted by the AEMM approach and the hog values given on the
Drawings. The actual measured hog is quite different, and not predicted by
the theory.

The AEMM approach has been applied to the beam with three different
assumptions of passive reinforcement in order to gauge to effect of the
passive reinforcement. Ignoring the passive reinforcement will result in over-
estimates of the beam hog. If the 2Y36 bars in the bottom of the beam are
included in the analysis, the upward hogs are reduced by almost 10mm. If the
full complement of passive reinforcement is included in the analysis, the effect
of the 8Y12 bars in the top flange of the beam tend to modify the influence of
the passive reinforcement to give a hog between the previous two values.

Two other cases have been analysed to consider the effect of hypothetically
increasing the passive reinforcement in the bottom of the beam to 6Y36 bars
in an extreme example. The effect of this large amount of passive
reinforcement is to reduce beam hogs and effectively stabilise their deflection
giving no increase in hog over time. Note that the initial deflections for these
cases are lower because the transformed section of the additional passive
reinforcement increases the moment of inertia of the total section.

3.0 DIFFERENTIAL SHRINKAGE AND CREEP IN THE CROSS


SECTION
The author, Connal (1997), has previously suggested that differential
shrinkage in Super T beams can result in some downward deflection in

Austroads 2006 – Deflections of Pretensioned Beams - Connal Page 6 of 10


beams. It was suggested that differential shrinkage alone can cause a
downward deflection of 10 to 20mm.

The top flange and webs of Super T beams are thinner than the bottom flange
zone which accommodates the prestressing strand. These thinner elements,
while in their non-composite state (i.e. prior to the deck being cast), have a
hypothetical thickness that is smaller than the thicker bottom flange zone. It
can be expected therefore that differential shrinkage strains could develop
through the depth of the beam. In the same way there can be differential
creep strains through the cross section for the same reason.

In normal circumstances, precast pretensioned beams are cast, the prestress


transferred and soon thereafter the beams are incorporated into the bridge by
making them composite with the deck casting. It is usual to assume in design
that the beam is about 1 month old when the deck is cast. However this can
of course vary depending on the construction circumstances. When the deck
is cast, the beam becomes composite with the deck slab and at the same time
additional downward sustain loading is added. At this point the effects of
creep and shrinkage are modified and reduced. The hypothetical thickness of
the beam components is also changed and the differential creep and
shrinkage effects are strongly modified.

As indicated above, the effect of differential shrinkage strains on beam


deflection has been considered previously [Connal (1997)] and has been
determined to be relatively small. AS5100 provides no warning of these
effects nor does it include any requirements to consider them. It is usual for
beams to be erected, and the deck to be cast, relatively quickly because this
reduces the overall construction period for the project. Usually the beams are
not left exposed to drying and ageing influences such that their behaviour is
significantly affected. In fact in most cases, precast beams keep hogging
upward and this forces their bearings to be adjusted downward to keep the
bridge surface profile in the design position (resulting in extra deck concrete at
the ends of the span). It is conceivable that if the beams are left for a long
time, in drying and exposed positions, differential creep and shrinkage can
occur. If for example, the beams are exposed to the summer sun and in
windy, drying conditions, with the top flange outstands shading the bottom
sections of the beam, the worst differential ageing conditions could be
envisaged for the beams.

The effect of both differential creep and shrinkage superimposed with the
effect of passive reinforcement, has been investigated for the case of the
beams of Span 1 of the Victorian bridge example, using the AEMM approach
indicated above. This is achieved by breaking the beam cross section into
zones and assigning varying hypothetical thicknesses to them, which in turn
defines different creep and shrinkage functions for their time dependent
behaviour.

The results of this analysis are indicated in Table 3. The AEMM computations
here are based on hypothetical thicknesses of 72mm for the top flanges,
100mm for the webs and 230mm for the bottom flange zone. This is a very

Austroads 2006 – Deflections of Pretensioned Beams - Connal Page 7 of 10


severe zoning of hypothetical thickness and has been combined with an
ambient relative humidity of 67%. Actual Bureau of Meteorology information
for relative humidity at the particular bridge site indicated an average summer
relative humidity of less than 60%. However, 67% relative humidity has been
adopted in the analysis, such that combined with the severe zoning of the
cross section, a realistic drying profile is adopted for what is a single beam
cross section. The three different sections of the beam cross section have
different creep and shrinkage functions, and the creep and shrinkage strains
they undergo, vary. This differential creep and shrinkage is accommodated
by the AEMM numerical analysis and the results are indicated below.

Table 3 — Deflections predicted by the AEMM Approach,


Including the Effects of Differential Creep and Shrinkage

Deflection (mm)

Time (days) 0 30 60 90 120


Ignoring passive reinforcement 58 37 37 38 39
With all passive reinforcement 55 33 32 33 34
(8Y12 top bars, 2Y36 bottom bars)
With only bottom passive reinforcement 54 29 27 28 28
(2Y36 bottom bars)
With increased passive reinforcement 48 20 16 15 15
(8Y12 top bars and 6Y36 bottom bars)
With only increased bottom passive 47 16 12 10 9
reinforcement (6Y36 bottom bars)

120 As-designed plus 30%

Indicated on drawings
100
As-designed

80 As-designed minus 30%


Deflection (mm)

AEMM ignoring passive


reinforcement
60
AEMM with all passive
reinforcement (8Y12 top bars, 2Y36
bottom bars)
40 AEMM with only bottom passive
reinforcement (2Y36 bottom bars)

Measured

20
AEMM with hypothetical increase
in passive reo (8Y12 top bars and
6Y36 bottom bars)
0 AEMM with hypothetical increase
in passive reinforcement (6Y36
0 30 60 90 120 bottom bars)

Time (days)

Figure 2 Deflection Response - “As-designed” & “Measured” versus “AEMM


Calculated” - AEMM calculations including differential hypothetical thicknesses

The plotted results from this analysis show agreement in the trend of the
measured deflection and the AEMM approach when the effects of differential
creep and shrinkage are included in the analysis. The predicted deflection

Austroads 2006 – Deflections of Pretensioned Beams - Connal Page 8 of 10


(represented by the ‘AEMM Top – 8Y12, Bottom – 2Y36’ curve) is about
10mm less than the measured value of deflection but the trend is very similar.

These results show that the deflections of beams are influenced more by the
effect of differential creep and shrinkage through the beam cross section than
the effect of load shedding to passive reinforcement. Some noteworthy
observations from this analysis are:

• The influence of the passive reinforcement in reducing hog is still present


when considered in conjunction with the effects of differential creep and
shrinkage, however the effects are somewhat modified.

• When the differential creep and shrinkage effects are considered in the
most severe circumstances (as assumed in the analysis) there is a
tendency for beams to deflect downwards with time, after the transfer of
the prestress. This behaviour is predicted irrespective of the amount of
passive reinforcement included, however increases with the quantity of the
passive tensile reinforcement.

4.0 CONCLUSIONS

It is difficult to be definitive from this investigation of these two examples of


beam sagging behaviour, because there are so many variables that affect
their behaviour. In particular the analysis of their behaviour is not precise, nor
is the modelling of the properties of the concrete that comprise them.
However there are several conclusions to be drawn:

• In both cases examined, the beams were exposed over a relatively long
period of time to relatively severe drying conditions. It is likely that these
conditions have led to differential creep and shrinkage effects that
combined with the moderate amount of passive reinforcement in the
beams and the inherent variability of the concrete, has resulted in low
beam hogs and beam sagging at service after incorporation in the bridge.

• The Age-adjusted Effective Modulus Method (AEMM) is an effective


modelling method for manual calculation of the deformation of
pretensioned beams. It can incorporate the influence of passive steel and
different shrinkage and creep behaviour in the different portions of the
beams.

• The use of passive (non-prestressed) reinforcing steel in pretensioned


beams is an effective means of providing strength and controlling beam
deflections, particularly in beams that would otherwise require large
amounts of pretensioned steel to achieve adequate strength. However
passive steel can significantly modify pretensioned beam deflections, if not
used moderately.

• For low levels of sustained loading (ie. when a beam is only subject to its
own self weight), increasing the amount of passive reinforcement in the
tensile zone (ie. in the bottom of the beam), decreases the negative or

Austroads 2006 – Deflections of Pretensioned Beams - Connal Page 9 of 10


hogging curvature with time, and can actually make the beam deflect
downwards with time if the amount of non-prestressed reinforcement is
sufficiently large.

• Beam cross sections with large variations of member thickness that allow
differing rates and amounts of creep and shrinkage strains to be
experienced, can cause significant deformation of the beam. Under
severe exposure conditions, the differential affects of creep and shrinkage
within the beam cross section can be large. This is not an effect that is a
normal design consideration nor is it an effect that the bridge design code,
AS5100, foreshadows or for which it provides any rules or guidance.

• Consideration should be given to the protection of pretensioned beams


from prolonged exposure to extreme drying conditions, particularly when
the beams have proportions that result in an asymmetric distribution of
significantly different element thicknesses. The case of Super T beams
exposed to severe summer drying conditions is an example.

5.0 REFERENCES AND BIBLIOGRAPHY

1. AUSTROADS, (1996) Australian Bridge Design Code (HB 77)

2. Connal, J.K., (1997) Bridge Beam Developments – the Teeroff, Austroads Bridging the
Millennia Conference, Sydney.

3. Gilbert, R.I., (1988) Time Effects in Concrete Structures, Elsevier, Amsterdam

4. Standards Australia, (2004), AS5100 Bridge Design

Austroads 2006 – Deflections of Pretensioned Beams - Connal Page 10 of 10

You might also like