See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/292548415

Practical determination of prestress tendon profile by load-balancing method

Article  in  Transactions Hong Kong Institution of Engineers · September 2006

DOI: 10.1080/1023697X.2006.10668050

CITATION READS

1 1,961

2 authors:

P.L. Ng Albert Kwan

The University of Hong Kong The University of Hong Kong
106 PUBLICATIONS   289 CITATIONS    275 PUBLICATIONS   3,820 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Sustainable Fibre-Reinforced Polymer Concrete Structures View project

All content following this page was uploaded by P.L. Ng on 08 July 2016.

The user has requested enhancement of the downloaded file.

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer
House, 37-41 Mortimer Street, London W1T 3JH, UK

HKIE Transactions
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/thie20

Practical Determination of Prestress Tendon Profile

by Load-balancing Method
a b
P L Ng BEng & Albert K H Kwan BSc(Eng) PhD CEng FHKIE MICE
a
Department of Civil Engineering,, The University of Hong Kong, E-mail:
b
Department of Civil Engineering,, The University of Hong Kong, E-mail:

To cite this article: P L Ng BEng & Albert K H Kwan BSc(Eng) PhD CEng FHKIE MICE (2006): Practical Determination of
Prestress Tendon Profile by Load-balancing Method, HKIE Transactions, 13:3, 27-35

To link to this article: http://dx.doi.org/10.1080/1023697X.2006.10668050

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to
anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss, actions,
claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or
indirectly in connection with or arising out of the use of this material.
 Practical Determination of Prestress Tendon
Profile by Load-balancing Method
P L NG BEng Albert K H KWAN BSc(Eng) PhD CEng FHKIE MICE
Department of Civil Engineering, The University of Hong Kong Department of Civil Engineering, The University of Hong Kong
Email: [email protected] Email: [email protected]

The determination of prestress tendon profile is a crucial step in the design of prestressed concrete structures. Like other
design procedures, it is perceived to be a trial-and-error process. As an uncommon alternative to the conventional method
of working out secondary moments and revising the tendon profile when the line of pressure falls out of the limiting
zone, the load-balancing method offers great potential for direct determination of tendon profile. Though decades have
passed since its proposition, there are still no comprehensive guidelines as to how the load-balancing method should
be implemented. Issues regarding how much load should be balanced and how the method should be applied to two-
dimensional structures etc have not been systemically addressed. In this paper, the load-balancing method is revisited.
Various considerations in the implementation of the method are discussed and two examples are presented to illustrate the
step-by-step procedures. It is shown that the method is actually quite easy to implement, even in complicated structures like
curved continuous bridges. Finally, it is advocated that the method is a much more efficient alternative to the conventional
method and should be incorporated in the standard design process.
Keywords: Load-balancing Method, Prestressed Concrete

Introduction
In the design of prestressed concrete structures, designers often found
the deviation of the line of pressure in concrete (C-line) from the tendon
profile due to structural indeterminacy incomprehensible. For example,
consider the two-span continuous beam in Fig 1(a) with a prestress tendon
running from one end to the other. Due to structural indeterminacy,
there are often vertical reactions induced at the supports by the prestress
forces, as shown in Fig 1(b). These reactions are called secondary loads.
They are self-equilibrating but nonetheless would produce secondary
moments, as depicted in Fig 1(c). Since the secondary moments are
produced by vertical reactions at supports, the secondary moment
diagram is linear. Due to the secondary moments, additional eccentricity
is introduced causing the Cline to deviate from the tendon profile,
Fig 1(d). The amount of deviation is linear and in fact proportional to
the secondary moment diagram.
Theories for the above phenomenon have been developed [1]. The first
theory is that of concordant profile. If the tendon profile produces no
secondary moment such that the C-line coincides with the tendon profile,
the profile is called a concordant profile; otherwise, it is a non-concordant
profile. The second theory is that of linear transformation. If the tendon
profile is linearly transformed by raising or lowering the tendon at the
intermediate supports while maintaining the shape of the tendon profile
between supports, there will be no effect on the position of the C-line.
Hence, the amount of deviation of the C-line from a non-concordant profile
may be taken as the linear transformation to be applied for obtaining a
concordant profile, and the C-line itself is a concordant profile.
In conventional practice, the following steps, which have been given in
many textbooks [1,2], are generally adopted to determine the tendon
profile and the C-line: (1) assume a trial tendon profile; (2) for the given
tendon profile, calculate the primary moments ignoring the secondary
loads; (3) calculate the resulting moments taking into account the
secondary loads; (4) evaluate the secondary moments as the difference
between the resulting moments and the primary moments; (5) apply
linear transformation to the tendon profile to yield the C-line; (6) if
the C-line does not lie entirely within the limiting zone governed by
the permissible stresses of concrete, modify the tendon profile and
repeat the above steps until the C-line lies entirely within the limiting
zone. Such a procedure to determine the tendon profile is basically a
trial-and-error process. It could be very iterative if the initially assumed
trial tendon profile is not close to any viable one. Good experience and Figure 1 – A Two-span Continuous Prestressed Concrete Beam

27
TRANSACTIONS • Volume 13 Number 3
judgement is therefore needed when assuming the trial tendon profile. 8Fh
For complicated structures, the determination of tendon profiles following wb =
L2
the above procedures could be a formidable task.
Whilst the primary moments can be evaluated quite easily because they Adding the external load and the equivalent load of the prestress force
are dependent only on the prestress force and eccentricity of the tendon, together, the net loading on the beam is only of intensity (w – wb). The
the computation of the secondary moments is rather cumbersome. beam needs only to carry the unbalanced load (w – wb), and the resulting
Conventionally, the equivalent load approach [2] of converting the prestress flexural stresses and deflection of the beam should be much smaller than
force to equivalent point loads, distributed loads and bending moments, those without prestressing. Furthermore, if the prestress force and tendon
and applying the equivalent loads to the statically indeterminate structure profile are designed such that wb = w, there would be no unbalanced
is used to compute the resulting moments and secondary moments. load and the beam would be subjected only to axial compression, which
Kong [3] has recently developed an alternative approach of using the eliminates tension and avoids cracking of the concrete.
influence lines due to a moving unit load and a moving unit moment to The load-balancing method can also be applied to two-dimensional
directly compute the secondary moments at the supports. His approach structures. Consider, for example, a two-way slab subjected to downward
has the major advantage that for structures comprising solely of linear load within the slab area. To balance the external load, we can provide
members, closed-form solutions for the secondary moments can be tendons in two orthogonal directions. Each set of tendons in one direction
obtained. The closed-form solutions would allow direct evaluation of would produce an upward equivalent load. If the prestress forces and
how the secondary moments would vary with the change of tendon tendon profiles are designed such that the total equivalent loads of the
profile and thus easier judgement of how the tendon profile should be tendons in the two directions more or less balance the external load,
modified when necessary. the net loading on the slab could be very much reduced.
Regardless of whether the conventional equivalent load approach or Kong’s The above method of balancing the external loads by prestressing so as
influence line approach is adopted, the determination of the tendon profile to minimise the flexural/shear stresses and deflection of the structure is
remains a trial-and-error process requiring the computation of secondary applicable to all kinds of structures, regardless of whether the structure
moments at each iteration step. Moreover, all the theories about secondary is statically determinate or indeterminate and whether the tendon profile
moments are applicable only to one-dimensional structures (structures is concordant or non-concordant. Most important of all, the secondary
comprising solely of linear members). For two-dimensional structures, loads induced by prestressing, which could produce not only secondary
such as a relatively wide bridge constructed of several beam members moments, but also secondary warping shears and secondary torsions, do
and a curved bridge sitting on several supports, the secondary loads not come into play and therefore need not be explicitly considered when
induced at the supports would produce not only secondary moments, applying the load-balancing method. It is necessary only to determine
but also secondary warping shears (loads causing warping of sectional the unbalanced loads and apply the unbalanced loads to the structure
shape) and secondary torsions (loads causing twisting). Extension of the to find out whether the resulting flexural/shear stresses and deflection
secondary moment theories to two-dimensional structures is extremely are within the acceptable limits.
difficult and would in any case result in procedures too complicated for
design applications.
In actual fact, the C-lines of the structure can be evaluated just by
analysing the statically indeterminate structure under the equivalent loads
of the prestress force without resorting to any secondary moments at all.
It is therefore advocated that we should do away with the calculation
of secondary moments during the iterative process of determining the
tendon profile. We cannot avoid searching for a viable tendon profile
iteratively but if the initially assumed tendon profile is close enough to
a viable one, then the number of iterations required should be relatively
small. Hence, the most important step is to find a good trial tendon
profile as the initially assumed tendon profile to start with. For this
Figure 2 – Balancing of External Load by Prestressing
purpose, the load-balancing method should be the most direct method,
as depicted in this paper.
Load-balancing Criteria
Load-balancing Method Apart from dead load (DL), there is also live load (LL) acting on the
structure. As the live load could vary greatly in both magnitude and
Load-balancing Concept position, many different loading cases need to be considered. However,
The concept behind the load-balancing method, which was originated it is possible only to balance a given set of external loads and therefore
by Lin [4] in 1963, is basically to conceive prestressing as a means of the presence of unbalanced load is inevitable, whichever loading case
balancing the external loads acting on the structure. When applying is considered for load balancing.
the method, the prestress force is replaced by equivalent loads along While emphasising the importance of judgement, Lin [4] has provided
the tendon. The equivalent loads comprise of point loads at sharp turns the following rules for choosing the loads to be balanced. Firstly, for
of the tendon, distributed loads at curves of the tendon and bending structures subjected to live load that is small compared to the dead
moments at ends of the tendon. If the prestress force and tendon profile load, we may balance the dead load only. Secondly, for structures
are designed such that the equivalent loads just balance or nearly subjected to live load that is not small compared to the dead load, we
balance the external loads, then there would be a relatively small net need to balance not only the dead load but also a certain portion of
loading on the structure, leading to reduced flexural/shear stresses and the live load. The basic criterion is that the resulting unbalanced loads
lateral deflection. in the different loading cases should not produce any flexural stresses
Take the simply supported prestressed concrete beam in Fig 2 subjected exceeding the allowable limits. If the amount of live load balanced is on
to a downward uniformly distributed load of intensity w as an example. the high side, then when no live load is acting, there will be a relatively
Let the prestress force be F, the span length be L and the sag of the large upward unbalanced load. On the other hand, if the amount of
tendon be h. If the tendon profile has a uniform curvature, ie the tendon live load balanced is on the low side, then when full live load is acting,
profile is parabolic, then the equivalent load of the prestress force would there will be a relatively large downward unbalanced load. In between,
be an upward uniformly distributed load of intensity wb per unit length there is an optimum point at which the upward unbalanced load and
given by: the downward unbalanced load would produce more or less the same

28 The Hong Kong Institution of Engineers

flexural stresses. For such optimisation, Lin suggested that the load to [7], where the quasi-permanent value is expressed as ψ2 times the live
be balanced might be taken as (DL + ½ LL). load. Hence, the value of α should not be greater than the value of ψ2
Aalami and Jurgens [5] have recently advised that for the design of given in these codes.
prestressed concrete floor slabs and beams, the load to be balanced may Without further ado, the value of α may in practice be simply assigned
be expressed in terms of the dead load only. They recommended that an initial value of 1/3 for a Class 1 structure and an initial value of 0
for floor slabs, 0.6 to 0.8 of the dead load could be balanced while for for a Class 2 structure. These values of α should each be close to the
floor beams, which are more heavily loaded, the load to be balanced optimum value for the respective class of structure in usual cases. When
could be increased to 0.8 to 1.1 of the dead load. No explanation was found necessary, the value of α may be adjusted upwards or downwards
given as to how these recommended values were derived. Presumably, to satisfy all the permissible stress limits at different stages or just to
these values were arrived at based on their own practical experience. optimise the design of the structure.
Other than the above, there have been very few guidelines for establishing
Determination of Tendon Profile
the load-balancing criteria. In this regard, the authors are of the view that
the loadbalancing criteria should be developed based on the following The load-balancing method offers great potential in simplifying the
considerations. Firstly, the load to be balanced should better be expressed structural design and analysis procedures. Notwithstanding this, the
in terms of both the dead load and the live load. In general, the load method has not gained wide recognition, especially in the local practice.
to be balanced may be taken as (DL + αLL), where α is a coefficient. One probable reason is the lack of step-by-step procedures for designers
The live load LL should be the live load of the critical loading case. If to follow. For several decades, the load-balancing method has not
there are several critical loading cases, then all these loading cases been further advanced in the mainstream development of prestressed
should be considered by working out the moment envelop produced concrete structures. In most textbooks on prestressed concrete design,
and evaluating the live load that would produce a moment diagram the load-balancing concept is only briefly mentioned [8] or even not
similar to the moment envelop as the live load LL to be considered included [9]. Hence, before this method would gain wider recognition
for load balancing. For instance, in the case of bridge design, several and eventually become a handy tool for design, more explanations and
loading cases with live load acting on different spans could be critical. guidelines need to be provided.
When considering the hogging moment at a support, the loading case
Some step-by-step procedures for the determination of tendon profile
of adjacent spans loaded would be critical but when considering the
are given in the following: (1) establish the load-balancing criteria and
sagging moment at mid-span, the loading case of alternate spans loaded
compute the loads to be balanced; (2) determine the tendon sag in each
would be critical. These two loading cases would together produce a
span and draft the idealised tendon profile in the form of discrete curves;
moment envelop with a maximum hogging moment at the support and
(3) taking the extrema of the idealised tendon profile as control points,
a maximum sagging moment at mid-span. From these, the live load LL
produce a trial tendon profile in the form of continuous curves; (4) check
may be taken as the live load that would produce a moment diagram
whether the extreme fibre stresses have exceeded the permissible stress
similar to the moment envelop. However, since the structure should be
limits in any loading case to verify the viability of the tendon profile;
able to carry some unbalanced load, there is no need to balance all
and (5) if the tendon profile is not viable, modify the tendon profile and
live load in all loading cases. To avoid becoming too sophisticated, the
repeat the above steps until a viable tendon profile is arrived at. With
live load LL to be considered for load balancing may simply be taken
the trial tendon profile guided by balancing a certain combination of
as the full live load acting on all spans.
loads, the above procedures should be more efficient than blind trials.
The next step is to choose an appropriate value of α, which is dependent Though iterations may still be needed, the number of iterations to arrive
on a number of factors. If α is too high, then at transfer stage when the at a viable tendon profile should be relatively small. The above steps
structure is subjected to only part of the dead load and no live load, are demonstrated by the following examples.
the upward unbalanced load may become critical, while if α is too low,
then at full load, the downward unbalanced load may become critical Examples
instead. Hence, the permissible stresses at transfer stage would impose
an upper limit to α and the permissible stresses at working stage would Example 1
impose a lower limit to α. In case the permissible stresses at transfer Example 1 is a straight three-span continuous bridge having a uniform
stage are exceeded, the value of α should be adjusted downwards. On cross-section in the form of a box girder. The middle span is 30 m and
the other hand, in case the permissible stresses at working stage are the two side spans are both 20 m. The dead and live loads acting on
exceeded, the value of α should be adjusted upwards. the bridge are 41.5 kN/m and 21.1 kN/m, respectively. Denoting the
The amount of load to be balanced is to some extent also dependent on four supports by A, B, C and D, a schematic diagram of the bridge is
the class of the prestressed concrete structure. For a Class 1 structure, shown in Fig 3. The bridge is to be post-tensioned by two tendons, one
in which no tension is allowed, a larger load generally needs to be embedded in each web, and designed as a Class 2 prestressed concrete
balanced and the appropriate value of α is likely to be higher, while structure. The concrete is of Grade 50.
for a Class 2 structure, in which some tension is allowed, a smaller Since the bridge is a Class 2 structure, we set α = 0 and balance only
load generally needs to be balanced and the appropriate value of α is the dead load by prestressing. To provide adequate concrete cover, the
likely to be lower. minimum vertical distance between the centreline of the tendon and the
Another important factor to be considered is that of excessive camber. In upper/lower flange surfaces is designed as 0.120 m. With this minimum
many structures, the live loads are essentially transient loads, each with vertical distance, for the middle span BC, the tendon sag is worked out
an average value substantially lower than the design value (the design as hBC = 1.100 – 2 x 0.120 = 0.860 m, from which the prestress force
value is actually a conservative estimate of the maximum possible value). is obtained as:
Consequently, the equivalent load of the prestress force could exceed
the total load, leading to an upward unbalanced load acting on the ⎛ 1 ⎞ ⎛ w bL 2 ⎞ (41.5)(30)2
F=⎜ ⎟⎜ ⎟= = 2714 kN
structure, during most of the time. The upward unbalanced load would ⎝ 2 ⎠⎝ 8hBC ⎠ (2)(8)(0.860)
cause cambering, ie upward deflection, of the structure. Such cambering
would increase with time due to the aggravating effect of creep and With this prestress force, for the side span AB, the tendon sag is
might become excessive in the longer term. To avoid excessive camber, evaluated as:
it is suggested that the live load to be balanced should not be greater
than the average value of the live load. For this purpose, the average ⎛ 1 ⎞ ⎛ w bL 2 ⎞ (41.5)(20)2
value may be taken as the quasi-permanent value of the live load hAB = ⎜ ⎟⎜ ⎟ = = 0.382 m
⎝ 2 ⎠ ⎝ 8F ⎠ (2)(8)(2714)
given in the Eurocode 0 [6] and the UK National Annex for Eurocode 0

29
TRANSACTIONS • Volume 13 Number 3
Figure 3 – Example 1: Bridge Elevation and Cross-section

To withstand the required prestress force, we may for each tendon use
22 x φ12.7 mm strands. Such a tendon has a minimum breaking load
of 4041 kN.
Having determined the respective tendon sag in each span, an idealised
tendon profile can be drafted in the form of discrete parabolic curves, as
shown in Fig 4(a). However, such an idealised tendon profile has sharp
cusps at the supports. To avoid kinking, the idealised tendon profile,
which consists of a series of discrete parabolic curves with gradient
discontinuity at the cusps, needs to be adjusted to become a continuous
curve with gradient continuity throughout. For making such adjustment,
the following method, which is an improved version of the method
developed by Lin and Burns [1], may be used. Firstly, control points
are established at the local maximum and minimum of the idealised
tendon profile. Then, a point of inflection is established at each side
of a cusp at a certain horizontal distance from the cusp. Each point of
inflection sets out the location where the curvature of the tendon would Figure 4 – Example 1: Provision of Prestress
reverse. Finally, a parabolic curve is fitted within each segment between
a control point and a point of inflection in such a way that the gradient condition is employed to determine the vertical position of F. The formulas
is equal to zero at the control point and the gradient is continuous at for determining the vertical position of F are given in Appendix II, from
the point of inflection. Details of the above process are explained in which it can be seen that to achieve gradient continuity at the point of
the following paragraphs. inflection, F must be collinear with EB. Similarly, the point of inflection
G divides the segment BH into two sub-segments. Two parabolic curves,
In this example, the local maximum of the idealised tendon profile occurs
one for each sub-segment, are fitted in such a way that the gradient is
at A and B while the local minimum occurs at E and H, as marked in
equal to zero at the control points B and H and gradient continuity is
Fig 4(b). The positions of the local minimum E and H may be evaluated
achieved at the point of inflection G. The trial tendon profile so formed,
using the formulas given in Appendix I. It should be noted that because
which is a continuous curve with gradient continuity throughout, is
of asymmetry, E is not at mid-span of AB and because of symmetry,
drawn in Fig 4(b).
H is at mid-span of BC (Lin and Burns have wrongly assumed that the
local minimum would always occur at mid-span). The points A, E, B and The equivalent loads of the prestress force with the idealised tendon
H are taken as control points. To smooth out the sharp cusp at B, we profile adjusted to become the smooth trial tendon profile are shown
establish a point of inflection F at the left side of B and another point in Fig 4(c). These equivalent loads are worked out using the following
of inflection G at the right side of B. The horizontal positions of F and equations, in which w AE , w E F , w FB , w BG and w G H are the equivalent
G are fixed by locating F at a horizontal distance from B equal to 1/10 distributed loads of the prestress force within the segments AE, EF, FB,
of the span length AB and locating G at a horizontal distance from B BG and GH, respectively:
equal to 1/10 of the span length BC (the distance of 1/10 span length
is set arbitrarily; other fractions, such as 1/8 or 1/12, of the span length (2)(8)(2714)(0.226)
may also be adopted to fix the horizontal positions of F and G). At this wAE = – = –41.6 kN/m
(2 x 7.683)2
stage, it should be borne in mind that the vertical positions of F and G
are yet to be fixed.
The point of inflection F divides the segment EB into two sub-segments (2)(8)(2714)(0.226 + 0.260)
wEF = – = –49.6 kN/m
EF and FB. For the sub-segment EF, a parabolic curve ending at E and F
(2 x 10.317)2
is fitted in such a way that the gradient is equal to zero at the control
point E. Likewise, for the sub-segment FB, a parabolic curve ending at
F and B is fitted in such a way that the gradient is equal to zero at the (2)(8)(2714)(0.354 – 0.260)
control point B. The two parabolic curves EF and FB meet at the point wFB = = 255.1 kN/m
of inflection F to form a curve EFB. To ensure smoothness, the curve (2 x 2.0)2
EFB must satisfy the condition that the gradient is continuous at F. This

30 The Hong Kong Institution of Engineers

 (2)(8)(2714)(0.354 – 0.182) Example 2
wBG = = 207.5 kN/m
(2 x 3.0) 2
This example is to demonstrate the application of the load-balancing
method to two-dimensional structures, in particular curved continuous
(2)(8)(2714)(0.182 + 0.506) bridges. The load-balancing procedures applied to two-dimensional
wGH = – = –51.9 kN/m structures are basically the same as those applied to one-dimensional
(2 x 12.0)2
structures. Hence, the load-balancing procedures remain relatively simple
regardless of the structural form. On the other hand, the conventional
Applying the above equivalent loads to the structure, the bending moments method becomes fairly complicated when applied to two-dimensional
at the critical locations E, B and H due to prestressing are found to be structures. For this reason, the simplifications to the design process
–1016 kNm, 2496 kNm and –2170 kNm, respectively. rendered by using the load-balancing method instead of the conventional
method are greater in two-dimensional structures. In fact, when dealing
In the next step, the bending moments due to prestressing are combined
with two-dimensional structures, the conventional theories about secondary
with those due to the external loads to evaluate the net bending moments
moments are not applicable at all. Therefore, it should be better to do
induced at the various critical locations. By placing the dead load on all
away with the calculation of secondary moments and perform structural
spans and the live load on appropriate span(s), the maximum bending
analysis directly to evaluate the stress conditions of the structure under
moments due to the external loads are calculated and plotted in the
the external loads and the equivalent loads of the prestress forces.
form of a moment envelop in Fig 5. From this figure, which also lists out
Taking a step further, when establishing the trial tendon profile, even the
the bending moment values at E, B and H under different loading cases
structural analysis can be dispensed with. It will be seen in this example
with the critical values underlined, the critical bending moments at E,
that if the load-balancing method is used, the trial tendon profile can
B and H due to the external loads are obtained as 1767 kNm, –4347
be established from the known loading pattern of the structure without
kNm and 3156 kNm, respectively. Putting the effects of external load
performing any structural analysis.
and prestressing together, we can compute the net bending moments
at E, B and H as: Another major advantage of employing the load-balancing method in the
design of two-dimensional structures is that we can utilise prestressing to
Mnet,E = 1767 – 1016 = 751 kNm balance not only the bending moments but also the torsions due to the
external loads. In a wide bridge, torsions would occur when the vertical
Mnet,B = –4347 + 2496 = –1851 kNm loads are larger at one side of the bridge than the other. Due to the
same cause, in a curved bridge, torsions would almost certainly occur
Mnet,H = 3156 – 2170 = 986 kNm because the outer rim is longer and heavier than the inner rim. If the
prestressing is designed in such a way that the pattern of the equivalent
loads due to the prestress forces would match the pattern of the external
loads (ie, where the external loads are larger, the prestressing is designed
to provide larger equivalent loads to balance the external loads), then
the net loading on the structure would produce much smaller torsions
than without prestressing.
The example is a curved multi-span highway bridge. It provides a two-
lane carriageway of total width 6.4 m and has the structural form of a
single cell box girder, as shown in Fig 6. The span along the longitudinal
axis measures 35.0 m. Due to the curved geometry, the span measured

Figure 5 – Example 1: Moment Envelop Due to External Load

From the above, it is obvious that the net bending moment at B is

governing. The concrete stresses at B under the loading case of dead
load on all spans plus live load on spans AB and BC are checked in
the following:
At top:

F Mnet,B 2 x 2714 x 103 –1851 x 10 6

fB = – – =– – = 0.38 MPa
A Zt 1.578 x 10 6 0.4842 x 10 9

At bottom:

F Mnet,B 2 x 2714 x 103 –1851 x 10 6

fB = – + =– + = –7.68 MPa
A Zb 1.578 x 10 6 0.4360 x 10 9

For a Class 2 post-tensioned member cast of Grade 50 concrete, the

allowable tensile stress is 2.6 MPa > 0.38 MPa, while the allowable
compressive stress is 19.8 MPa > 7.68 MPa. Hence, the trial tendon
profile is viable. No iterative modification of the tendon profile until it
is viable is needed. Figure 6 – Example 2: Bridge Plan and Cross-section

31
TRANSACTIONS • Volume 13 Number 3
along the outer web is 36.3 m while that along the inner web is 33.7
m. Such span geometry is repeated for the adjacent spans at both ends.
The self-weight of the box girder is 95.5 kN/m. In addition, the weight of
the road surfacing is 14.1 kN/m and the weight of the parapet/signage
at each side of the bridge is 3.0 kN/m. The traffic load on the bridge is
HA loading, which consists of a uniformly distributed load of 30.0 kN/m
per lane and a knife-edge load of 120.0 kN per lane. The box girder is
to be post-tensioned by four tendons, two in each web, and designed
as a Class 1 structure. The concrete is of Grade 50.
As the bridge is a Class 1 structure, we set α = 1/3 and balance all the
dead load plus 1/3 of the full live load (full live load is the live load
when all lanes are loaded) by prestressing. The external loads on the
outer side of the longitudinal axis of the bridge are applied to the outer
web while the external loads on the inner side of the longitudinal axis
are applied to the inner web. Assuming that the weights of the girder
and road surfacing are uniformly distributed over the deck surface and
the weight of the parapet/signage is a uniformly distributed line load
acting at the edge of the bridge, the dead load acting on the outer web
is (95.5 + 14.1)/2 + 3.0 x (37.7/36.3) = 57.9 kN/m and the dead load
on the inner web is (95.5 + 14.1)/2 + 3.0 x (32.2/33.7) = 57.7 kN/m.
On the other hand, assuming that all lanes are loaded and smearing
the knife-edge load over the span, the live load acting on each web is
approximately 30.0 + 120.0/35.0 = 33.4 kN/m. Adding the dead load
with 1/3 of the live load, the external loads to be balanced by the outer
and inner webs are 57.9 + 33.4/3 = 69.0 kN/m and 57.7 + 33.4/3 =
68.8 kN/m, respectively.
As for other curved bridges, the total external loads to be balanced by
prestressing of the outer web are heavier than that to be balanced by
prestressing of the inner web. This is because the outer web is longer
and the deck area supported by the outer web is larger. However, in
order to avoid producing any lateral bending or movement of the deck Figure 7 – Example 2: Idealised Tendon Profiles
in the horizontal plane, the prestress forces in the outer and inner webs
should be set equal (if the prestress forces in the two webs are not the 2 x 0.200 – 0.296 = 0.954 m. The idealised tendon layout so derived
same, there will be bending moment about the vertical axis of the cross- for the inner web is shown in Fig 7(b). To withstand the required prestress
section). The difference in the equivalent loads of the prestress forces in force, we may for each tendon use 44 x φ12.7 mm strands. This tendon
the outer and inner webs has to be achieved by adjusting the geometry, has a minimum breaking load of 7715 kN.
mainly the sag, of the tendons in the two webs.
The next step is to smoothen the idealised tendon profiles in Fig 7
To provide adequate concrete cover, the minimum vertical distance to generate the trial tendon profiles using the same procedures as in
between the centreline of the tendon and the upper/lower flange Example 1. The trial tendon profiles so generated are displayed in Fig
surfaces is set as 0.140 m. Furthermore, to avoid local crushing of the 8. With these trial tendon profiles, the equivalent loads of the prestress
concrete between adjacent tendons in the same plane of curvature, the forces are evaluated as per the following equations, in which wAD,out ,
minimum vertical distance between the centrelines of the two tendons wDC,out , wAD,in and wDC,in are the equivalent loads of the prestress forces
in each web is set as 0.260 m. With these minimum vertical distances, within the segments AD in the outer web, DC in the outer web, AD in
for the outer web, the tendon sag is worked out as hout = 1.650 – the inner web and DC in the inner web, respectively:
2 x 0.140 – 0.260 = 1.110 m, which is applicable to both tendons in
the web. The idealised tendon profile so derived for the outer web is
shown in Fig 7(a). Having determined the tendon sag, the prestress force (2)(8)(5119)(0.410 – 0.188)
wAD,out = = 345.0 kN/m
of each tendon in the outer web is obtained as: (0.2 x 36.3)2

⎛ 1 ⎞ ⎛ w bL 2 ⎞ (69.0)(36.3)2
F= ⎜ ⎟⎜ ⎟ = = 5119kN (2)(8)(5119)(0.188 + 0.700)
⎝ 2 ⎠ ⎝ 8hout ⎠ (2)(8)(1.110) wDC,out = – = –86.2 kN/m
(0.8 x 36.3)2
As explained before, the prestress force of each tendon in the inner web
is set equal to that in the outer web. With this same prestress force, for (2)(8)(5119)(0.332 – 0.141)
wAD,in = = 344.4 kN/m
the inner web, the tendon sag is evaluated as:
(0.2 x 33.7)2

⎛ 1 ⎞ ⎛ w bL 2 ⎞ (68.8)(33.7)2
hin = ⎜ ⎟⎜ ⎟= = 0.954 m (2)(8)(5119)(0.141 + 0.622)
⎝ 2 ⎠ ⎝ 8F ⎠ (2)(8)(5119) wDC,in = – = –86.0 kN/m
(0.8 x 33.7)2
Since the tendon sag for the inner web is smaller, the clear distances
between the tendon and the upper/lower flange surfaces and between
adjacent tendons in the same plane of curvature may be increased Finally, the equivalent loads of the prestress forces in the outer and inner
to allow easier construction. To achieve this required tendon sag, the webs are combined with the external loads to evaluate the net loading
vertical distance between the centreline of the tendon and the upper/ on the structure. By placing the dead load on all spans, the live load on
lower flange surface may be set as 0.200 m and the vertical distance appropriate lane(s) and span(s), and the equivalent loads of the prestress
between the centrelines of the two tendons in the web may be set as forces at their respective locations, the maximum bending moments due
0.296 m. Such vertical distances would yield a tendon sag of 1.650 – to the net loading are evaluated by grillage analysis and plotted in the

32 The Hong Kong Institution of Engineers

 Figure 9 – Example 2: Moment Envelop Due to Net Loading

Conclusions
The conventional method of determining the tendon profile based on the
concept of secondary moments and the load-balancing method based
on the concept of balancing part of the external loads by prestressing
have been reviewed, and their pros and cons studied. Between them,
the conventional method is more popular. However, it is not really a
good design method for two-dimensional structures simply because in
such structures the secondary loads due to prestressing produce not only
secondary moments but also secondary warping shears and torsions.
On the other hand, due probably to the lack of design guidelines, the
load-balancing method is not commonly used. In actual fact, the load-
balancing method has the major advantages of allowing straightforward
Figure 8 – Example 2: Trial Tendon Profiles appreciation of the goal of prestressing and direct determination of the
trial tendon profile.
form of a moment envelop in Fig 9. The maximum bending moments
To facilitate systematic implementation of the load-balancing method, an
at the support A and the mid-span location C due to the net loading
account of the various considerations for establishing the load-balancing
are summarised in the following:
criteria, some step-by-step procedures for direct determination of the
trial tendon profile, and an improved method of converting the idealised
Mnet,A,out = –4317 kNm
tendon profile containing sharp cusps into a continuous tendon profile
with gradient continuity throughout have been presented. Two examples
Mnet,A,in = –3356 kNm
are given to demonstrate the application of the method to one- and
two-dimensional structures. Through these examples, it is shown that the
Mnet,C,out = 3076 kNm
load-balancing method is equally applicable to one- and two-dimensional
structures and can greatly simplify the determination of tendon profile
Mnet,C,in = 2057 kNm
especially in the case of two-dimensional structures.
It is seen that the net bending moment acting on the outer web at Finally, it is advocated that the load-balancing method is a much
support A is governing. The concrete stresses in the critical section (ie more efficient alternative to the conventional method and should be
the outer web at A) are checked in the following: incorporated in the standard design process. The step-by-step procedures
presented herein and the improved method of deriving the continuous
At top:
tendon profile given in the appendices can be easily computerised
to automate the process of determining the prestress tendon profile.
F Mnet,A 2 x 5119 x 103 –4317 x 10 6
fA = – – =– – = –0.75 MPa Furthermore, the load-balancing method is not limited to any kind of
A Zt 0.5 x 3.8995 x 10 6
0.5 x 1.9168 x 10 9 structures. It should be applicable also to a wide range of structures,
including slabs, frames, shells and folded plates etc.

At bottom:

F Mnet,A 2 x 5119 x 103 –4317 x 10 6 References

fB = – + =– + = –11.67 MPa
A Zb 0.5 x 3.8995 x 10 6 0.5 x 1.3454 x 10 9
1. Lin, T.Y. and Burns, N.H., Design of Prestressed Concrete, 3rd Edition, pp646,
John Wiley and Sons, New York, USA (1981).
2. Nawy, E.G., Prestressed Concrete: A Fundamental Approach, 5th Edition, pp945,
Tension does not occur anywhere in the critical section. Hence, the no Prentice Hall, New Jersey, USA (2006).
tension requirement of a Class 1 structure is satisfied. For a post-tensioned 3. Kong, J., A practical method for the calculation of secondary prestress moments,
member cast of Grade 50 concrete, the allowable compressive stress HKIE Transactions, Vol 11, No 4, pp49-52 (2004).
4. Lin, T.Y., Load balancing method for design and analysis of prestressed concrete
is 19.8 MPa > 11.67 MPa. It is therefore verified that the trial tendon
structures, ACI Journal, Vol 60, No 6, pp719-742 (1963).
profiles for the tendons in the outer and inner webs are viable. The 5. Aalami, B.O. and Jurgens, J.D., Guidelines for the design of post-tensioned
tendon profiles have been arrived at directly without any trial-and-error floors, Concrete International, Vol 25, No 3, pp77-83 (2003).
iteration.

33
TRANSACTIONS • Volume 13 Number 3
6. British Standards Institution, BS EN 1990: 2002: Eurocode – Basis of Structural Appendix II
Design (formerly known as Eurocode 0), British Standards Institution, UK,
(2002).
Point of Inflection on Trial Tendon Profile
7. British Standards Institution, NA to BS EN 1990: 2002: UK National Annex
for Eurocode 0 – Basis of Structural Design, British Standards Institution, UK,
(2004). Consider the segment of idealised tendon profile shown in Fig A2. It
8. Rajagopalan, N., Prestressed Concrete, 2nd Edition, pp486, Alpha Science, starts at the local minimum M1 and ends at the local maximum M2.
Harrow, UK (2005). This segment of idealised tendon profile is to be replaced by two reverse
9. Kong, F.K. and Evans, R.H., Reinforced and Prestressed Concrete, 3rd Edition, parabolas joined together at a point of inflection P to form a continuous
pp508, Van Nostrand Reinhold (UK) Co. Ltd., Berkshire, UK (1987). curve. From M1 to P, a parabolic curve is to be fitted in such a way that
the gradient is equal to zero at M1. From P to M2, another parabolic
curve is to be fitted in such a way that the gradient is equal to zero at
M2. Let the coordinates of M1, M2 and P be (xM1 , yM1), (xM2 , yM2 ) and
Appendix I (xP , yP ), respectively. The equation of the curve fitted between M1 and
P can be derived as:
Local Minimum of Idealised Tendon Profile

Consider the idealised tendon profile shown in Fig A1. It is given in the yP – yM1
y – yM1 = (x – xM1)2
form of a parabolic curve linking the first end point with coordinate (xP – xM1)2
(x1 , y1) to the second end point with coordinate (x2 , y2). Let the horizontal
distance between the two end points (given by x2 − x1 and same as Differentiating, the gradient of this curve at P may be obtained as:
the span length) be L, the vertical distance between the two end points
(given by y2 − y1) be Δy, and the tendon sag be h. The equation of the dy 2(yP – yM1)
=
parabolic curve can be derived as:
dx x=xP (xP – xM1)

⎛ 4h ⎞ ⎛ Δy ⎞
y – y1 = ⎜ ⎟ (x – x1)(x – x2) + ⎜ ⎟ (x – x1) Likewise, the equation of the curve fitted between P and M2 can be
⎝ L2 ⎠ ⎝ L ⎠ derived as:

Differentiating with respect to x, the gradient of the curve may be yP – yM2

obtained as: y – yM2 = (x – xM2)2
(xP – xM2)2
dy ⎛ 4h ⎞ ⎛ Δy ⎞
= ⎜ ⎟ (2x – x1 – x2) + ⎜ ⎟ Differentiating, the gradient of this curve at P may be obtained as:
dx ⎝ L 2 ⎠ ⎝ L ⎠
dy 2(yP – yM2)
The local minimum occurs at where the gradient is equal to zero. Setting =
dx x=xP (xP – xM2)
the above equation for the gradient to zero and solving for x, the
x-ordinate xM of the local minimum is evaluated as:
Equating the gradients of the two curves at P, the following equation
⎛ L ⎞⎛ Δy ⎞ for yP is derived:
xM = x1 + ⎜ ⎟ ⎜1 – ⎟
⎝ 2 ⎠⎝ 4h ⎠ yP – yM1 yM2 – yM1
=
Substituting the value of xM into the equation of the parabolic curve, the xP – xM1 xM2 – xM1
y-ordinate yM of the local minimum is evaluated as:
Hence, the point of inflection P must be collinear with M1 and M2.
2
⎛ Δy ⎞
yM = y1 – h ⎜ 1 – ⎟
⎝ 4h ⎠

Figure A2 – Point of Inflection on Trial Tendon Profile

Figure A1 – Local Minimum of Idealised Tendon Profile

34 The Hong Kong Institution of Engineers

 P L NG Albert K H KWAN
P L Ng is currently a PhD student in the Department Albert K H Kwan is a Professor and an Associate Dean
of Civil Engineering, the University of Hong Kong. He of Engineering at the University of Hong Kong. He
obtained his first degree in civil engineering in 2003 has worked on a number of major civil engineering
at the same university. His research interests include projects before pursuing an academic career. Since
nonlinear finite element analysis of reinforced concrete joining the University of Hong Kong in 1989, he has
structures and thermal, shrinkage and time-dependent been actively researching on concrete technology, tall
movement analysis of concrete structures. building structures and earthquake engineering. He was
recently awarded the Telford Premium by the Institution
of Civil Engineers in UK, and the HKIE Transaction Prize
and the Special Award in Research Project by the Hong
Kong Institution of Engineers.

35
TRANSACTIONS • Volume 13 Number 3

View publication stats