Engineering Structures 35 (2012) 203–214

Contents lists available at SciVerse ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Collapse mechanisms and strength prediction of reinforced concrete pile caps

Uffe G. Jensen 1, Linh C. Hoang ⇑
Institute of Technology and Innovation, University of Southern Denmark, Campusvej 55, 5230 Odense, Denmark

a r t i c l e i n f o a b s t r a c t

Article history: This paper describes an upper bound plasticity approach for strength prediction of reinforced concrete
Received 1 June 2011 pile caps. A number of collapse mechanisms are identified and analysed. The procedure leads to an esti-
Revised 29 August 2011 mate of the load-carrying capacity and an identification of the critical collapse mechanism. Calculations
Accepted 2 November 2011
have been compared with nearly 200 test results found in the literature. Satisfactory agreement has been
Available online 7 January 2012
found. The analyses are conducted on concentrically loaded caps supported by four piles. The paper
briefly outlines how the approach may be extended to more complicated loadings and geometries. It is
Keywords:
argued that the upper bound approach may be a useful complement to the widely used lower bound
Collapse mechanisms
Reinforced concrete
strut-and-tie method. Especially when dealing with strength assessment of existing structures.
Concrete plasticity Ó 2011 Elsevier Ltd. All rights reserved.
Pile caps
Bridge foundations

1. Introduction 3-dimensional models which may be used in practice. Even though

This paper deals with strength prediction of reinforced concrete
pile caps and has been prepared on the basis of a Ph.D. project con-
ducted by the first author, [1]. A pile cap is basically a foundation
block supported by piles. In high rise buildings, pile caps are for
example used to support columns while in bridge engineering, pile
caps are typically used to transfer vertical and transverse loads
from the bridge pier to the piles, see illustration in Fig. 1. Depend-
ing on the applied loads and the number of piles, pile caps can be
rather massive structures. Typically, the centre to centre distance
between adjacent piles is about three times the pile diameter
and the cap thickness is larger than the pile diameter.
Strength prediction of pile caps has traditionally been based on
a sectional approach where the sectional forces are calculated
based on beam theory [2,3]. Pile caps, however, are normally char-
acterised by a small span to depth ratio where it is easier for the
shear action to be transmitted directly to the supports. For that
reason the more recent design codes, e.g. Eurocode [4] and AASTHO
LRFD [5], recommend that strength verification of pile caps may be
based on strut-and-tie models.
Strut-and-tie modelling has in recent years dominated the
research into limit analysis of pile caps [6–11]. The method is
based on the lower bound theorem of plastic theory and therefore
provides lower bound solutions for the load-carrying capacity. A
major challenge in strut-and-tie modelling is to develop suitable
⇑ Corresponding author. Tel.: +4565507372; fax: +4565507364.
E-mail addresses: ugj@cowi.dk (U.G. Jensen), lch@iti.sdu.dk (L.C. Hoang).
1
Present address: COWI A/S, Parallelvej 2, 2800 Kgs. Lyngby, Denmark.
the approach is conceptually simple, 3-dimensional models may
become quite complex, especially in the nodal zones. It is therefore
not unusual to see pile caps designed by use of plane strut-and-tie
models in practice. This is a very simplified and conservative ap-
proach, which may be acceptable in design situations but is less
suitable when dealing with strength assessment of existing
structures.
When assessing the strength of existing structures, engineers
often strive to determine both lower bounds as well as upper
bounds for the load-carrying capacity. This is an efficient way to
narrow down the interval within which the actual load-carrying
capacity can be expected and thus a way to decide whether the
structure needs strengthening or not. For pile caps, lower bound
solutions may – as mentioned – be obtained from strut-and-tie
models. Investigations of upper bound methods for pile caps have
not been reported in the literature.
The objective of this paper is to demonstrate how an upper
bound plasticity approach may be used to predict the critical fail-
ure mode and the load-carrying capacity of pile caps. The basic idea
is to identify and analyse a number of collapse mechanisms and
take the lowest calculated capacity as an estimate for the actual
load-carrying capacity. Failure in pile caps is normally divided into
three main categories, namely punching shear failure, shear failure
and flexural failure. These distinct failure modes are investigated in
this paper. Local anchorage failure is not considered as such failure
should in practice be prevented by adequate detailing of the rein-
forcement. The investigations reveal that punching failure is not
likely to occur for typical pile cap geometries. The calculations
have been compared with nearly 200 test results and satisfactory
agreement has been found.

0141-0296/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2011.11.006
204 U.G. Jensen, L.C. Hoang / Engineering Structures 35 (2012) 203–214

Notation

Roman letters Pu,f flexural capacity

a vector Pu,p punching shear capacity
a1 shear span Pu,s sectional shear capacity
b vector T sectional shear force in pile
ba distance from centre of pile to centre of column WE external work
bc width of column WI internal work
c vector WI,c internal work (dissipation) in concrete
c =½l  ba  ½D WI,s internal work (dissipation) in reinforcement
d depth of pile cap
de effective depth Greek letters
fc uniaxial cylinder compressive strength of concrete a angle between failure surface and the displacement vec-
ft tensile strength of concrete tor
fy yield stress of reinforcement a1, a2 angles between failure surface and the displacement
k H/P vector
l side length of pile cap b angle defining inclination of failure plane
n1, n2 normal vectors u angle of internal friction for concrete (tan u = 3=4 )
u (ux, uy, uz) vector describing relative displacement in k1, k2 constant
failure surface m effectiveness factor
u displacement m0 factor taking into account microcracking and softening
w crack opening ms crack sliding reduction factor
Ai area of surface Si (i = 1, 2, . . ., 6) h angle of rotation
As reinforcement area per unit length ql longitudinal reinforcement ratio
D diameter of pile sexp Pexp/(2dl)
H transverse sectional shear force su Pu/(2dl)
M sectional moment su,f Pu,f/(2dl)
Mpile sectional moment in pile su,p Pu,p/(2dl)
P load on pile cap su,s Pu,s/(2dl)
Pexp experimental ultimate load U mechanical degree of reinforcement
Pk=0 Pu for k = 0
Pu calculated load-carrying capacity

the four piles configuration also forms the basis for the present
investigation. However, as discussed in Section 2.4, it is rather
straight forward to extend the upper bound approach to pile caps
with more than four piles and pile caps with eccentric loadings.
It is assumed that the pile cap is provided with an orthogonal
reinforcement mesh at the bottom face. The reinforcement areas
per unit length in both directions are assumed to be identical
and termed As. The reinforcement material is assumed to be rigid,
perfectly-plastic with yield stress fy. Furthermore, it is assumed
that the reinforcement is well-anchored, so that local anchorage
failure does not take place.
The concrete is treated as a Modified Coulomb material obeying
the normality condition of plastic theory and having an angle of
internal friction u given by tan u = 3/4, see Nielsen and Hoang
[20]. The tensile strength of concrete is neglected and the plastic
compression strength is taken as mfc where m is the so-called effec-
tiveness factor (see later) and fc is the uniaxial cylinder compres-
sive strength.

2.1. Punching shear failure

Punching shear failure in a massive concrete block like a pile

Fig. 1. Illustration of bridge substructure with pier, pile cap and piles.
2. The approach
Most research concerning pile caps has been focused on four-
pile caps with concentric loading [6–19]. To enable comparisons,
cap might be difficult to imagine. In plastic analyses of punching
shear in slabs, it is usually assumed that the longitudinal reinforce-
ment does not yield. This implies that the relative displacement u
in the failure surface (yield line) must be perpendicular to the plan
of the slab (see Fig. 2). According to the normality condition of
plastic theory, the angle a between u and the failure surface cannot
be smaller than the angle of friction, i.e. a P u is required. With
this restriction and with the assumption of u being perpendicular
to the plane of the slab, a geometrically admissible punching
 U.G. Jensen, L.C. Hoang / Engineering Structures 35 (2012) 203–214 205

(a)

Fig. 2. Punching shear failure in slab with large span to depth ratio.

mechanism can only be developed if the supports are placed far

away from the concentrated punching load. The theoretical plastic
solution for this pure punching case may be found in [20].
For deep members such as pile caps, it is obvious that a punch-
ing mechanism as outlined is not possible. The relatively small
span to depth ratio would not allow a ‘‘pure’’ punching mode. In
order to fulfil the condition a P u, the failure mode has to be a
combination of a downward punch of the central part and a hori-
(b)
zontal displacement of the remaining parts of the cap, see Fig. 3.
This implies that the reinforcement has to be in a state of yielding.
Furthermore, in addition to the inclined punching failure surface, a
number of vertical failure planes must be formed when the
remaining parts are horizontally displaced.
Bearing the above discussions in mind, the following punching
mechanism is considered (see Fig. 4 and the list of notation for
explanation of symbols). A central concrete block is punched
through the pile cap by the column load P. Due to symmetry, this
block is subjected to a downward displacement parallel to P. The
displacement increment (or rate) of the central part may be writ-
ten as [dux, duy, duz] = k1[0, 0, 1], where k1 is a positive constant.
The shape of the punched concrete block is described by a number
of inclined planes. These planes are denoted S1, S2, S4 and S5 in
Fig. 4a and b. Owing to symmetry, S1 is similar to S4 and S2 is sim-
ilar to S5. In order to fulfil a P u, the remaining of the pile cap is
assumed to be split into four parts (I, II, III and IV) where each part
is horizontally displaced. As indicated in Fig. 4a, part I is separated
from parts II and IV by the vertical planes S3 and S6. These planes
are due to symmetry subjected to pure separation failures, i.e.
the relative displacement here is perpendicular to the failure plane.
The displacement increment for part I may be written as [dux, duy,
duz] = k2[1, 1, 0] where k2 is a positive constant. The corresponding (c)
displacement increments for parts II, III, IV are k2[1, 1, 0],
k2[1, 1, 0] and k2[1, 1, 0] respectively. Due to symmetry, only
the failure planes confining part I need to be considered.

Fig. 4. Punching shear mechanism in pile cap: (a) 3D view; (b) plan view;
(c) diagonal section B–B.

Instead of using k1 and k2 (or more correctly the ratio k1/k2) to

define the relative displacement between part I and the central
Fig. 3. Punching shear failure in slab with small span to depth ratio. part, it is also possible to describe the relative displacement by
206 U.G. Jensen, L.C. Hoang / Engineering Structures 35 (2012) 203–214

the vector u and the angle a as shown in Fig. 4b and c. By calculat- (see illustration in Fig. 5). The term |u| is the norm of the relative
ing the external work and the dissipated energy, the work equation displacement u. It should be noticed that a1 and a2 are functions
can be applied to establish an upper bound solution for P. of the angle a defined in Fig. 4c.
Denoting the areas of the planes S1, S2, S4 and S5 as A1, A2, A4 and As shown in Fig. 5a and b, the plane S1 is described by the vec-
A5 respectively, the total area of the punching failure surface can be tors a and b while S2 is described by the vectors a and c. The angles
written as 4(A1 + A2 + A4 + A5) = 8A1 + 8A2. The energy, WI,c, dissi- a1 and a2 may be obtained from
pated in these concrete failure planes can be found by applying
juS1 j ju  ðu  n1 Þ  n1 j
the dissipation formula for Modified Coulomb materials as given cos a1 ¼ ¼ ;
juj juj
in [20]. The result is
juS2 j ju  ðu  n2 Þ  n2 j
cos a2 ¼ ¼ ð2Þ
1 juj juj
W I;c ¼ mfc ½ð1  sin a1 Þ8A1 þ ð1  sin a2 Þ8A2 juj ð1Þ
2 where n1 = (a/|a|)  (b/|b|) is the normal to S1 and n2 = (c/|c|) 
(a/|a|) is the normal to S2. The surface areas A1 and A2 can be
Note that there is no contribution to WI,c from S3 and S6 because
expressed as |a  b|/2 and |a  c|/2, respectively. The vectors a, b,
these planes undergo pure separation failure and ft = 0 has been
c are given by (see geometrical definitions in Fig. 4a):
assumed. In Eq. (1) a1 and a2 are the angles between u and its pffiffi
21 3 2 3 2 3
projections uS1 and uS2 on the failure planes S1 and S2, respectively
2
ðl  bc Þ ba  12 bc  4
2
D 0
6 7 6 pffiffi 7 6 1 7
a ¼ 4  12 bc 5; b ¼ 4 ba  1 bc  2
D 5; c ¼ 4  2 bc 5 ð3Þ
2 4
d d 0
(a) The angle b shown in Figs. 4c and 5 can be determined from:
d
tan b ¼ pffiffiffi pffiffi ð4Þ
2ba  22 bc  12 D
The reinforcement is in a state of yielding at the positions where it
crosses the concrete failure planes, see illustration in Fig. 6. For one
quarter of the plane section shown in Fig. 6, the internal work per-
formed by the yielding reinforcement is equal to ½lAs fyux + ½lAs fyuy.
Here ux and uy are the first and second component in the displace-
ment vector u. The total dissipation in the reinforcement, WI,s, may
be written as:
l
W I;s ¼ 4As fy ðux þ uy Þ ð5Þ
2
The external work, WE, can be found as the punching force P multi-
plied by the vertical component uz of the displacement vector u.
Now, by expressing u as follows:

(b)

Fig. 5. Definition of vectors and angles related to: (a) plane S1; (b) plane S2. Fig. 6. Plane section in pile cap at reinforcement level.
 U.G. Jensen, L.C. Hoang / Engineering Structures 35 (2012) 203–214 207

2 3 2 pffiffi2 3
ux  2 cosða þ bÞ mechanism). The part to the left of the yield line gets a displace-
6 7 6 pffiffi2 7 ment u relative to the central part. The displacement vector forms
u ¼ 4 uy 5 ¼ 4 cosða þ bÞ 5juj ð6Þ
2 the angle a with the yield line.
uz sinða þ bÞ An upper bound solution, Pu,s, may be obtained by applying the
an upper bound solution, Pu,p, for the punching shear mechanism same procedure as above, i.e. establishing expressions for the
can be found when solving the work equation WE = WI,c + WI,s. The internal and external work and inserting them into the work equa-
solution is found to be tion. The upper bound solution, su,s, appears as follows:
pffiffiffi su;s Pu;s ð1  sin aÞ  2 Um sin b cosða þ bÞ
4mfc ½ð1  sin a1 ÞA1 þ ð1  sin a2 ÞA2   8lAs fy cosða þ bÞ ¼ ¼ ð10Þ
Pu;p ¼ mfc 2ldmfc 2 sin b sinða þ bÞ
sinða þ bÞ
ð7Þ Solution (10) is only valid (i.e. corresponds to an admissible failure
mechanism) when both conditions a P u and a + b P 90° are ful-
By introducing the average shear stress s and the mechanical filled. The first condition is, as already mentioned, due to the nor-
degree of reinforcement U defined below mality condition of plastic theory. The second condition is
imposed to avoid the situation where the mechanism implies com-
1
P As fy pressive yielding in the reinforcement. When minimising Eq. (10)
s¼2 ; U¼ ð8Þ
ld dfc with respect to a, an analytical solution, see Eq. (12a), is obtained.
This minimum solution is found when a is given by:
 
the solution (7) can be rewritten as follows: 2U
cosða þ bÞ ¼  1 sin b ð11Þ
su;p Pu;p m
¼
mfc 2ldmfc If a found from (11) does not satisfy the conditions a P u and
pffiffi
2 2U a + b P 90°, the minimum of Eq. (10) will not be a valid solution.
ld
½ð1  sin a1 ÞA1 þ ð1  sin a2 ÞA2   m cosða þ bÞ
¼ ð9Þ When Eq. (11) gives a < u, a valid solution is derived by inserting
sinða þ bÞ a = u into (10). In this way, Eq. (12b) is obtained. Finally, if Eq.
To obtain an optimal upper bound solution, Eq. (9) has to be mini- (11) gives a + b < 90°, then a must be chosen such that a + b = 90°.
mised with respect to a and subjected to the constraints a1 P u This leads to solution (12c). Since cot b = a1/d and tan u = 3=4 , it is
and a2 P u. This may be done numerically. immediately seen that the condition a P u will be violated before
condition a + b P 90° when a1/d < 3=4 . The opposite will be the case
2.2. Shear failure when a1/d > 3=4 . The complete solution therefore reads:
8 "sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi #
> 1 a 2 4U  U

a1
>
> 1 
Instead of punching failure, a more traditional shear mechanism >
> þ 1  ; if a P u and a þ b P 90 ðaÞ
>
>
>
2 d m m d
(i.e. a beam shear failure mode) may be more critical. In this case, a >
>   
diagonal yield line with the inclination b is assumed to run from su;s < 3  4 ad1 Um þ ad1 2 þ 1
¼
mfc > ; if a < u and ad1 < 34 ðbÞ
the edge of the piles to the edge of the column, see Fig. 7 (note that >
> " 3 ad1 þ 4
>
> rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi #
the angle b here is not identical to b defined for the punching >
> a 2
>1
> 1 a1 
>
: þ1 ; if a þ b < 90 and ad1 P 34 ðcÞ
2 d d
ð12Þ
It should be noticed that (12a) together with (12c) are identical to
the plastic shear strength solution for shallow, rectangular beams
published by Nielsen and Braestrup [21]. Nielsen and Braestrup ar-
rived at these equations by assuming plane stress conditions, which
means that only a + b P 90° needs to be maintained. For details, see
also [20].

2.3. Flexural failure

Pile caps are sometimes designed with very low degree of lon-
gitudinal reinforcement. This means that a flexural collapse may
govern the load-carrying capacity. Fig. 8 illustrates the assumed
flexural failure mechanism. The displacement of part I relative to
part II is described by a rotation h around the edge of the column.
The concrete failure plane (yield line) separating the two parts
undergoes pure separation failure. Thus, no energy is dissipated
in the concrete as ft = 0 has been assumed. Consequently, only
the yielding reinforcement contributes to the internal work.
By considering the pile reactions as the driving forces, the exter-
nal work and the internal work for half of the pile cap appear as
follows:
 
1 1
W E ¼ 2 P ba  bc h; W I ¼ As lfy de h ð13Þ
4 2
By applying the work equation and introducing s and U defined in
Eq. (8), an upper bound solution su,f is found to be

Fig. 7. Shear mechanism in pile cap.

208 U.G. Jensen, L.C. Hoang / Engineering Structures 35 (2012) 203–214
Fig. 8. Flexural mechanism in pile cap.
su;f Pu;f Ude
¼ ¼ ð14Þ
mfc 2ldmfc mðba  12 bc Þ
This upper bound solution will, of course, result in a larger capacity
than the flexural strength determined from a traditional cross sec-
tional analysis. This is so because the assumed collapse mode does
not account for the depth of the compression zone at the cross sec-
tion with maximum bending moment. However, for small rein-
forcement degrees, the concrete compression zone is small
compared to the depth d of the pile cap. Further, as mentioned by
Clarke [12], the concrete below the column will be in a state of tri-
axial compression which will further decrease the necessary com-
pression zone. Hence, for practical purposes the simple solution
(14) will only result in a negligible overestimate of the flexural
capacity. The difference becomes larger for larger reinforcement de-
grees. In this case, however, a shear failure will govern the load-car-
rying capacity. This can be seen in Section 2.4, where calculation
results are shown.
2.4. Procedure for strength evaluation
The governing failure mechanism will be the one leading to the
smallest ultimate strength. Hence, su must be taken as the mini-
mum of su,f, su,s and su,p. To quantify su,s and su,p, it is necessary
to have a formula for the effectiveness factor m. This factor is intro-
duced to account for the fact that concrete is not a perfectly plastic
material as assumed. A comprehensive discussion on the necessity
of the effectiveness factor when applying plastic theory to struc-
tural concrete has been given in [20]. Traditionally, m is determined
by calibrating the theoretical solutions with test results. This
means that the effectiveness factor depends on the problem con-
sidered. In [20], formulas for m have been given for a large number
of problems. For instance, the effectiveness factor is generally lar-
ger for punching problems in two way spanning slabs than for
shear problems in slender beams.
The question remains as to what the effectiveness factor should
be taken as, when dealing with pile caps? Recent research in con-
crete plasticity seems to indicate that for a number of related prob-
lems, a unified formula for the effectiveness factor can be used
[20]. Basically, both in punching as well as in ordinary shear fail-
ures, there is a sliding resistance in the yield lines (failure planes)
that has to be overcome before collapse take place. This sliding
resistance depends strongly on whether the failure takes place in
cracked or in uncracked concrete. The concept of crack sliding
was studied by Zhang [22] who established a model for shear
strength prediction of concrete beams without stirrups. To distin-
guish between sliding failure in cracked and uncracked concrete,
Zhang proposed the following effectiveness factor:
ms m0 ; cracked concrete
m¼ ð15Þ
m0 ; uncracked concrete
The idea behind this formula is to separate the effect of cracking
from the effect of softening. The factor ms is called the crack sliding
reduction factor and was introduced to reflect the reduced cohesion
in a crack compared to that of uncracked concrete. By calibration
with beam test results, Zhang [22] suggested a value of ms = 0.5.
Based on a micromechanical model [23], Zhang also carried out
analyses of push-off tests on specimens with pre-made crack and
found that the proposed value for ms leads to reasonable results in
this case too. The factor m0 appears as follows, [22]:
 
0:88 1
m0 ¼ pffiffiffiffi 1 þ pffiffiffi ð1 þ 26ql Þ ðfc in MPa; d in mÞ ð16Þ
fc d
Here, the dependency on fc and d reflect softening and thus size ef-
fect whereas the dependency on the longitudinal reinforcement ra-
tio ql is mainly attributable to dowel action. Eqs. (15) and (16) have
in recent years been found to be applicable to a wide range of re-
lated problems, including shear strength of lightly shear reinforced
beams, punching strength of slender slabs and shear strength of cir-
cular cross sections [20].
 U.G. Jensen, L.C. Hoang / Engineering Structures 35 (2012) 203–214
It seems natural to investigate whether Eqs. (15) and (16) can
be used for the problems studied in this paper. The idea is that
m = msm0 should be used when the collapse mechanism is composed
of failure planes (yield lines), which prior to the sliding failure can
be identified as potential crack planes. On the other hand, if the
failure planes considered cannot be identified as potential crack
planes prior to sliding, the factor m = m0 is used. To fulfil the first
mentioned condition, a cracking mechanism must exist. Indeed,
such a cracking mechanism is easy to imagine for the shear failure
mode considered in this paper. Fig. 9 illustrates the principle. Prior
to shear failure (i.e. sliding failure) the crack is developed by a fi-
nite rotation of part I. The crack opening has been illustrated in
Fig. 9 by the displacement field w perpendicular to the crack. After
cracking, the reinforcement becomes active and prevents further
rotation (note that the case of continuing rotation is covered by
the flexural failure mechanism treated above). The final failure
then takes place as a combination of sliding and separation in
the crack. This is illustrated in Fig. 9 by the constant displacement
field u. For the punching mechanism, it is not possible to develop a
corresponding cracking mechanism. The surface suffering sliding
failure (i.e. the combination of the planes S1, S2, S4 and S5 shown
in Fig. 4) is concave. This means that a cracking mechanism which
simultaneously allows all these planes to undergo a so-called mode
I failure (i.e. pure crack opening) cannot be developed. Conse-
quently, it cannot be justified to consider these planes as poten-
tial/existing crack planes prior to punching failure.
Based on the above discussions, it is assumed that punching
failure involves sliding in uncracked concrete while shear failure
involves sliding in cracks. Hence, the procedure for strength evalu-
ation is as summarised in Eq. (17).
Fig. 9. Crack opening followed by shear failure.
209
8
< su;f
> Eq: ð14Þ ðflexuralÞ
su ¼ min su;s Eq: ð12Þ ðshearÞ with m ¼ ms m0
>
:
su;p Eq: ð9Þ ðpunching shearÞ with m ¼ m0
ð17Þ
Here, su,p should be understood as the minimum of Eq. (9) with re-
spect to a. A parametric study of su has been carried out and the re-
sults are shown in Fig. 10(a–d). In the calculations the ratios D/d and
bc/d have been set equal to 0.4 and 0.8 for some typical a1/d-ratios.
The normalised strength, su/m0fc, has been plotted versus U/m0. The
calculations have been made by fixing the ratios of de/d and c/D to
0.9 and 0.2, respectively.
As can be seen in Fig. 10(a–d), the flexural failure mode is gov-
erning for low values of U/m0. For increasing values of U/m0, shear
failure will be critical. It appears that the transition point at which
punching failure becomes critical is found at rather high degrees of
reinforcement. In fact, such high degrees of reinforcement are
rarely seen in practice. It is noticed that this transition point is
shifted to higher values of U/m0 when the a1/d-ratio decreases. If
the ratio D/d is increased with bc/d, c/d and de/d unchanged, punch-
ing failure becomes critical for lower values of U/m0 whereas if bc/d
is increased with D/d, c/d and de/d unchanged, punching failure be-
comes critical for higher values of U/m0.
It appears from the calculations that in practice, where U/m0 is
typically smaller than 0.2, punching shear in pile caps is not a real-
istic failure mode. This is an important observation, because exten-
sions of the outlined procedure to other cases then become fairly
simple when only flexural and shear failure need to be considered.
For illustration, consider the eccentrically loaded pile cap shown in
Fig. 11. For any admissible distribution of pile reactions (including
transverse shear and bending moments), it will be straight forward
to determine the external and the internal work by considering the
pile reactions as the driving forces. Four possible planes of failure
have been indicated. When each one of these is investigated for
shear failure, both the horizontal and the vertical reactions of the
relevant piles will contribute to the external work. For flexural
mechanisms, the bending moments in the piles will also contrib-
ute. To illustrate the general procedure, calculations for an eccen-
tric loaded pile cap have been provided in the Appendix.
3. Comparison of tests and calculations
Experimental data on pile caps is limited and the majority of the
tests have been carried out with four-pile caps. The most common
reinforcement layouts in the tests are bunched square (the rein-
forcement is concentrated as bands over the piles) or grid rein-
forcement (uniform distributed mesh reinforcement). These two
layouts are also mostly used in practice. In this paper only tests
with the mentioned reinforcement layouts are compared with
calculations.
The first published tests on four-pile caps can be found in [18].
This series comprises 59 tests, out of which 30 specimens with
bunched square or grid reinforcement layout have been included
in the present investigation. In the studies by Clarke [12] 15 spec-
imens were tested and 13 of those were with bunched square or
grid reinforcement layout. A comprehensive test program was
conducted in Japan [13–17], where 170 specimens have been
tested. 144 specimens were designed with bunched square or grid
reinforcement layout. Finally, a small series of three tests has been
reported in [19].
In total 190 tests have been used to evaluate the proposed
calculation procedure for four-pile caps. Detailed information
regarding geometry, material properties, test results and observed
failure modes have been summarised in [1]. The calculated results
corresponding to the three considered failure mechanisms can
210 U.G. Jensen, L.C. Hoang / Engineering Structures 35 (2012) 203–214
Fig. 10. Theoretical strength su/m0fc versus U/m0.
Fig. 11. Potential yield lines in an eccentrically loaded pile cap critical to shear and
flexural.
additional be found in [1]. The minimum ultimate load in each case
has been used to compare with the observed capacity. In the calcu-
lations of m0, the reinforcement ratio ql has been determined by
including only the reinforcement bars placed directly over the
piles. This was done to reflect the fact that for the same total
amount of reinforcement, higher capacity can be obtained by the
bunched reinforcement layout than by the grid layout (see [7]).
The corresponding values of calculated and observed strength
have been plotted in Fig. 12. The mean value of the ratio sexp/su
is 1.01 and the standard deviation is 0.15. According to the calcu-
lations, shear failure is governing for 126 specimens. For this
group, the mean value of sexp/su is 1.03 with a standard deviation
of 0.17. Flexural failure is governing for 64 specimens. Here, the
mean value of sexp/su and the corresponding standard deviation
are 0.98 and 0.09, respectively. The agreement between tests and
calculations is judged to be satisfactory – especially when com-
pared with the agreement obtained by for example Souza et al.
[9]. They used the tests reported in [12–16,18] to calibrate a num-
ber of parameters in their strut-and-tie model. In this way, Souza
et al. were able to minimise the standard deviation to 0.23 and ob-
tained a mean value of sexp/su equal to 1.01. The mean value corre-
sponds to the value found by use of the present model. However,
this does not imply that the upper- and lower bound models have
provided exact plastic solutions. Rather, the result is coincidental
since the calibration parameters in the model by Souza et al. [9]
 U.G. Jensen, L.C. Hoang / Engineering Structures 35 (2012) 203–214 211

failure mode for reinforcement ratios normally encountered in

practice. The procedure will be illustrated in the following example
with an eccentric loaded pile cap. The main purpose is to demon-
strate how the external work alters as compared to the equations
derived for the concentric loading case.

A.1. Loads and pile reactions

Fig. 12. Comparison of calculated strength with test results.
are not related to the effectiveness factors used in the present
model.
According to the calculations, [1], punching shear failure is not
critical for any of the tests. The explanation can be found in the fact
that the maximum value of U/m0 in the tests is 0.318. As can be
seen from Fig. 10, this value has a sufficient margin to the level
of U/m0 required to make punching failure critical.
Consider the four-pile cap shown in Fig. A1. The cap supports a
short column with length L1. At the top of the column, a vertical
load P acts simultaneously with a horizontal load H. Proportional
loading with H = kP is assumed. In bridge engineering, it is com-
mon to have k in the order of 0.03–0.06. For simplicity, dead loads
are disregarded.
To assess the strength of the cap, calculations of the pile reac-
tions are required. For the purpose of this example, it is sufficient
with the following estimates, which have the same degree of de-
tails as in a preliminary design phase in practice. P and the over-
turning moment from H are assumed to be carried by axial
reactions, N1 and N2, in the piles. The horizontal drift of the cap
due to H induces transverse shear forces, T, and bending moments
in the piles. To determine the distribution of bending moments, it
is assumed that the piles are fully restrained by the cap. Further,
the piles are considered fully restrained at a distance L2 below
the cap. The moment distribution will then be linear with a magni-
tude Mpile at the restrained ends (see Fig. A1). From equilibrium

4. Conclusions

The paper has dealt with analysis of collapse mechanisms and

strength prediction of reinforced concrete four-pile caps. Calcula-
tions have been compared with nearly 200 test results. The ob-
tained agreement is judged to be satisfactory, especially when
considering the fact, that an existing formula for the effectiveness
factor has been adopted.
The analyses have revealed that punching shear failure is critical
only when the pile cap is heavily reinforced. This means that in
most practical cases, flexural failure or shear failure will be critical.
This observation has the important impact that the approach then –
in an operational way – can be extended to deal with eccentrically
loaded pile caps and piles caps supported by more than four piles.
In general, upper bound solutions are easier to develop than
lower bound solutions. For this reason, the outlined approach
may be a useful complement to the more widely used strut-and-
tie method – both in design situations as well as in strength assess-
ment of existing structures. When carrying out design, estimates of
the overall dimensions can efficiently be obtained from upper
bound calculations while local strut-and-tie models may be used
for reinforcement detailing. When dealing with existing structures,
upper bound calculations may supplement strut-and-tie models to
narrow down a qualified estimate of the load-carrying capacity.
To further explore the potential of upper bound calculations of
pile caps, more research attention is required. This includes analyt-
ical and experimental studies of more complicated loading and
geometrical configurations.

Appendix A. Example with eccentric loading

As discussed in Section 2.4, it is simple to extend the analysis

procedure because punching has been found to be an unrealistic Fig. A1. Pile cap subjected to eccentric loading.
212 U.G. Jensen, L.C. Hoang / Engineering Structures 35 (2012) 203–214

conditions, the following estimates of T, Mpile, N1 and N2 are The internal work may be calculated as follows:
obtained:
W I ¼ As lfy de h ðA:6Þ
1 1
T ¼ H ¼ kP ðA:1Þ Now, by inserting the right hand side of Eqs. (A.1)–(A.3) into Eq.
4 4
TL2 1 (A.5) and by solving WE = WI with respect to P, an upper bound solu-
M pile ¼ ¼ kPL2 ðA:2Þ tion, P(a), is obtained.
2 8

  As lfy de
N1 P M pile HðL1 þ dÞ 1 1 k  L2 kðL1 þ dÞ PðaÞ ¼    ðA:7Þ
¼   ¼P   1
þ 4bk a ðL2 þ 2L1 þ 2dÞ ba  12 bc  4k ð2d þ L2 Þ
N2 4 ba 4ba 4 8 ba 4ba 2

ðA:3Þ
A.2.2. Mechanism (b)
A.2. Analysis of failure mechanisms In mechanism (b) T and Mpile are assisting the rotation of part I.
Hence, when calculating WE, the terms Tdh and Mpileh are added to
Fig. A2 shows the failure mechanisms to be considered. Mecha- the work performed by N1:
nisms (a) and (b) are flexural failures and mechanisms (c) and (d)  
1
are shear failures. In all four cases, part I moves relative to part II W E ¼ 2ðN1 ba  bc þ Td þ M pile Þh ðA:8Þ
as indicated and the pile reactions are considered as the driving 2
forces. When all four mechanisms have been analysed, an estimate For this mechanism, the internal work may also be calculated by
of the load-carrying capacity may be determined as: use of Eq. (A.6). The upper bound solution, P(b), then appears as
Pu ¼ MinfPðaÞ ; PðbÞ ; P ðcÞ ; PðdÞ g ðA:4Þ follows:
(a) (b) (c) (d)
where P , P , P and P are (see below) the results found for the As lfy de
mechanisms (a), (b), (c) and (d), respectively. PðbÞ ¼    ðA:9Þ
1 k
2
 4ba
ðL2 þ 2L1 þ 2dÞ ba  12 bc þ 4k ð2d þ L2 Þ
A.2.1. Mechanism (a)
It is noticed that Eqs. (A.7) and (A.9) will be reduced to Eq. (14)
Consider the flexural mechanism (a). The external work is:
when k = 0 (i.e. the case of concentric loading).
1
W E ¼ 2ðN2 ðba  bc Þ  Td  M pile Þh ðA:5Þ
2 A.2.3. Mechanism (c)
This equation differs from the expression given in Eq. (13). First, N2 Consider now the shear mechanism (c). Here, N2 performs work
replaces P/4. Further, the terms Tdh and Mpileh are subtracted be- through the displacement component |u|sin(a + b). It is assumed
cause T and Mpile counteract the rotation of part I. that the cap geometry is compact, i.e. a1/d  0.2–0.4 (as indicated

(a) (b)

(c) (d)

Fig. A2. Flexural mechanisms (a) and (b) and shear mechanisms (c) and (d).
 U.G. Jensen, L.C. Hoang / Engineering Structures 35 (2012) 203–214
in Fig. A2). This implies that (a + b) > p/2 when the normality con-
dition a P u is imposed (see also the discussions in Section 2.2).
Hence, T acts in the opposite direction of the horizontal component
of u and thereby counteracts the displacement of part I. The bend-
ing moment Mpile does not perform work in this case as there is no
rotation. The following expression for WE is established:
W E ¼ 2ðN2 sinða þ bÞ  T cosðp  a  bÞÞjuj ðA:10Þ
The internal work is:
ld 1
WI ¼  mfc ð1  sin aÞjuj þ As lfy cosðp  a  bÞjuj ðA:11Þ
sin b 2
This expression is identical to the one used to derive Eq. (10). The
first term is the dissipation in the concrete failure surface (yield
line) and the second term is the dissipation in the reinforcement.
The term ½mfc(1  sin a)|u| is the dissipation per unit area (see
[20]) and the term ld/sin b is the area of the failure surface.
By inserting Eqs. (A.1)–(A.3) into Eq. (A.10) and by solving the
work equation, the following upper bound solution is obtained:
(a)
(b)
Fig. A3. Reduction of load-carrying capacity Pu versus k for data set: d = 1.2 m,
fy = 400 MPa, fc/fy = 0.1, c/d = 0.2, bc/d = 0.4, a1/d = 0.3, de/d = 0.9, L1/d = 3 and
L2/d = 6.
213
mfc ð1  sin aÞ sinld b  2As lfy cosða þ bÞ
PðcÞ ¼    ðA:12Þ
1 þ bka L22 þ L1 þ d sinða þ bÞ þ k cosðb þ aÞ
This solution must be minimised with respect to a and subjected to
the condition a P u. This may be done numerically. As discussed in
Section 2.4, m = msm0 when dealing with shear failures.
A.2.4. Mechanism (d)
For mechanism (d), WI is calculated by use of Eq. (A.11). The
external work in this case has positive contributions from both
N1 and T:
W E ¼ 2ðN1 sinða þ bÞ þ T cosðp  a  bÞÞjuj ðA:13Þ
(d)
The solution, P , to be minimised with respect to a and subjected
to the condition a P u appears as follows:
mfc ð1  sin aÞ sinld b  2As lfy cosða þ bÞ
PðdÞ ¼    ðA:14Þ
1  bka L22 þ L1 þ d sinða þ bÞ  k cosða þ bÞ
It is noticed that Eqs. (A.12) and (A.14) are reduced to Eq. (10) when
k = 0.
A.3. Numerical results
Some results obtained from the above procedure are shown in
Fig. A3. The calculations were performed by fixing the following
parameters: d = 1.2 m, c/d = 0.2, bc/d = 0.4, a1/d = 0.3, de/d = 0.9, L1/
d = 3, L2/d = 6, fy = 400 MPa and fc = 40 MPa.
The figure depicts ratios of Pu/Pk=0 versus k, where Pk=0 simply
means the load-carrying capacity in the case without horizontal
load (note that Pk=0 may also be calculated from Eqs. (12) and
(14)). The graph in Fig. A3(a) has been determined by assuming
D/d = 0.4 and As = 0.00654 m2/m. This reinforcement intensity
may for instance be obtained by placing pairs of Ø25 mm
rebars at spacing 150 mm. The results in Fig. A3(b) are valid for
D/d = 0.6 and As = 0.00196 m2/m (corresponding to Ø25 mm rebars
at spacing 250 mm).
It appears that the ultimate load Pu reduces substantially when
the system is subjected to combinations of vertical and horizontal
loads. For k  0.05, which is a typical value in practice, approxi-
mately 20–30% reduction of Pu is observed in the cases studied.
As indicated, mechanism (c) is critical for the case covered by
Fig. A3(a). In Fig. A3(b), a shift of critical failure mode is observed
at k = 0.032. In the beginning, the flexural mechanism (a) is critical.
However, for k > 0.032, the shear mechanism (c) becomes decisive.
References
[1] Jensen UG. Limit analysis of reinforced concrete bridge substructures – shear
strength of piers and piles with circular cross section and strength of pile caps.
Ph.D. thesis, Institute of Technology and Innovation, University of Southern
Denmark, Denmark; 2011.
[2] American Concrete Institute. Building Code Requirement for Structural
Concrete (ACI 318M-95) and Commentary (ACI 318RM-95). Detroit; 1995.
[3] British Standard Institution. BS5400 Steel, concrete and composite bridges.
Part 4: Code of practice for design of concrete bridges. London; 1990.
[4] European Committee for Standardisation. EN 1992-1-1, Eurocode 2: Design of
Concrete Structures. Part 1: General Rules and Rules for Buildings. CEN,
Brussels; 2005.
[5] American Association of State Highway and Transportation Official. AASHTO
LRFD Bridge Design Specifications. 4th ed. Washington DC; 2007.
[6] Siao WB. Strut-and-tie model for shear behaviour in deep beams and pile caps
failing in diagonal splitting. ACI Struct J 1993;90(4):356–63.
[7] Adebar P, Zhou LZ. Design of deep pile caps by strut-and-tie models. ACI Struct
J 1996;93(4):1–12.
[8] Leu LJ, Huang CW, Chen CS, Liao YP. Design of deep pile caps by strut-and-tie
models. J Struct Eng 2006;132(6):929–37.
[9] Souza R, Kuchma D, Park J, Bittencourt T. Adaptable strut-and-tie model for
design and verification of four-pile caps. ACI Struct J 2009;196(2):142–50.
214 U.G. Jensen, L.C. Hoang / Engineering Structures 35 (2012) 203–214
[10] Souza R, Kuchma D, Park J, Bittencourt T. Nonlinear finite element analysis of
four-pile caps supporting columns subjected to generic loading. Comp
Concrete 2007;4(5):363–76.
[11] Park J, Kuchma D, Souza R. Experimental study on bending ultimate strength of
four-pile caps. Can J Civ Eng 2008;35(12):1399–413.
[12] Clarke JL. Behaviour and design of pile caps with four piles. Cement and
Concrete Association, London, Technical Report No. 42.489, 1973.
[13] Suzuki K, Otsuki K. Experimental study on corner shear failure of pile caps.
Trans Jpn Concrete Inst 2002;23:303–10.
[14] Suzuki K, Otsuki K, Tsubata T. Influence of edge distance on failure mechanism
of pile caps. Trans Jpn Concrete Inst 2000;22:361–7.
[15] Suzuki K, Otsuki K, Tsubata T. Influence of bar arrangement on ultimate
strength of four-pile caps. Trans Jpn Concrete Inst 1999;21:327–34.
[16] Suzuki K, Otsuki K, Tsubata T. Experimental study on four-pile caps with taper.
Trans Jpn Concrete Inst 1998;20:195–202.
[17] Otsuki K, Suzuki K. Experimental study on bending ultimate strength of
four-pile caps. Trans Jpn Concrete Inst 1996;482:93–102.
[18] Blevot J, Frémy R. Semelles Sur Pieux. Annales de l’Institut Technique du
Batiment es des Travaux Publics 1967;20(230):223–95 (In French).
[19] Chan TK, Poh CK. Behaviour of precast reinforced concrete pile caps. Constr
Build Mater 2000;14:73–8.
[20] Nielsen MP, Hoang LC. Limit analysis and concrete plasticity. 3rd ed. CRC Press;
2011.
[21] Nielsen MP, Braestrup MW. Shear strength of prestressed concrete
beams without web reinforcement. Mag Concrete Res 1978;30(104):
119–28.
[22] Zhang JP. Diagonal cracking and shear strength of reinforced concrete beams.
Mag Concrete Res 1997;4(178):55–65.
[23] Zhang JP. Strength of cracked concrete. Part 2: Micromechanical modelling of
shear failure in cement paste and in concrete. Technical University of
Denmark, Department of Structural Engineering and Materials, Report No.
17, Lyngby, 1997.