Engineering Structures 167 (2018) 481–493

Contents lists available at ScienceDirect

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

Splitting force of bottle-shaped struts with different height-to-width ratios T

a a b c,⁎
Aimin Yuan , Donghui Xu , Shoulong Qian , Hang Dai
a
Dept. of Civil and Transportation Engineering, Hohai Univ., No. 1 Xikang Rd., Nanjing 210098, China
b
Shanghai Tonghao Civil Engineering Consulting Co. Ltd, 210000 Nanjing, China
c
College of Civil Engineering, SEU, No. 2 Sipailou, Nanjing 210096, China

A R T I C LE I N FO A B S T R A C T

Keywords: A bottle-shaped strut is a common kind of strut in the structures, whose compression disperses, resulting in a
Bottle-shaped struts splitting force at some distance away from the load plate. Considering characters of the geometric and physical
Splitting force boundary conditions which are different from previous studies, a new mathematical equation for the infinite
Compression-dispersion models isostatic lines of the compression (ILC) is obtained and a corresponding equation of transverse stresses dis-
Loading area ratios
tribution is derived. Then theoretical formulae to calculate the resultant splitting forces and their location in a
Height-to-width ratios
bottle-shaped strut are proposed. Based on the principle of superposition, different compression dispersion
models for the struts with different height-to-width ratios are individually formulated and simplified. A re-
commendation for EN1992 is presented. The transverse stresses distribution and the equations for the resultant
splitting force and their locations are validated by the test and FEM results. The results show that the proposed
equations have good agreements with the test and FEM result, which can provide accurate predictions for the
magnitude and location of the resultant splitting force.

1. Introduction
A strut and tie model consists of struts, ties and nodes [1]. It has
been thought as an effective tool for modelling disturbed regions in
structural concrete members [2]. A bottle-shaped strut is a common
kind of strut in the structures such as deep beam and pile caps. In this
type of strut, the load is applied to a relatively small area of the
member, which inevitably results in the compression stress dispersing
with a convex outer profile like a bottle as shown in Fig. 1. To maintain
static equilibrium, a transverse tensile force has to be developed to
balance the later component of the inclined compressive forces. Con-
sequently, as the applied load increases, those strains in the struts ex-
ceed the tensile resistance of concrete and a dominant splitting crack is
formed along the strut axis.
To date, efforts to build a better splitting force formula for a bottle-
shaped strut have been made by some researchers. The transverse
splitting tension in disturbed region (D-region) was discussed at first by
Guyou [3], who used isostatic lines of compression (ILC) to investigate
the dispersion of loads in post-tensioned anchorage zones. He creatively
sketched the basic configuration of a compression-dispersion model
(CDM). Sahoo et al. [4] presented a mathematical description of this
model for bottle-shaped struts and derived an equation for estimating
the splitting force. However, some researchers [5,6] think the theore-
tical model proposed by Sahoo is questionable. The transverse
compressive stress below the anchor plate is equal to zero, which is not
in a agreement with the result of FEA and Breen’s conclusion [7].
Therefore, He et al. [6] modified the physical boundary conditions and
proposed a revised CDM. Brown [8] calculated the splitting force in
terms of the slope of dispersion of compression and the efficiency
factor. Zhou [9] introduced the boundary condition of the three and
four order derivative of the equation of ILCs, and derived a sixth-order
polynomial equation, and then the bursting force formula was obtained.
Ghanei and Aghayari [10] investigated the effect of height-to-width
ratios on the dispersion of compression, and they found the location of
maximum tensile straining and the dispersion angle of compression are
dependent on the height-to-width ratios of the bottle-shaped strut. It
indicates that current ACI 318-14 recommending a 2:1 fixed dispersion
model may underestimate the splitting force of a bottle-shaped strut
whose height-to-width ratio (h/b) is less than 2 as shown in Fig. 1(a),
but suitable for a bottle-shaped strut whose height-to-width ratio (h/b)
is greater than 2 as shown in Fig. 1(b). In EN 1992-1-1 2004 [11], an
equation for full discontinuity subjected to h/b < 2 has already pro-
vided for calculating splitting forces in bottle shaped strut as follows:
1 a
Tb = ⎛1−1.4 ⎞ P
4⎝ h⎠ (1)
However, as the height-to-width ratio becomes smaller and smaller,
one equation cannot consider D-region superposition effect of the

⁎
Corresponding author.
E-mail addresses: yuanam@163.com (A. Yuan), 18260625043@163.com (D. Xu), shoulongqian@163.com (S. Qian), hansdai@sohu.com (H. Dai).

https://doi.org/10.1016/j.engstruct.2018.04.048
Received 22 May 2017; Received in revised form 12 April 2018; Accepted 13 April 2018
0141-0296/ © 2018 Elsevier Ltd. All rights reserved.
A. Yuan et al.
(a) h/b < 2 (b) h/b 2 search works, we still try to establish the polynomials equations of
Fig. 1. Isostatic lines of compression and STM model for bottle-shaped struts.
bottle-shaped struts under different scenarios.
Seen from reviewed literature mentioned above, the present re-
search hotspots still put focus on how to reasonably define the as-
sumptions of the geometric and physical boundary conditions of the
ILCs. They remain assuming the first order derivative of the equation of
ILC is equal to zero at x = 0 even if the load plate is considered in their
models. And no one considers the significance of the turn point where
the transverse stresses change from the tension to the compression. So
there is still controversy among different researchers, and further stu-
dies are still necessary. Furthermore, current theories do not cover all
conditions, especially for a bottle-shaped strut with height to width
ratio less than 2. The superposition effect should be considered to cal-
culate the splitting force of a bottle-shaped strut.
2. Research significance
In this paper, a new quantitative compression-dispersion model,
considering the different assumptions based on the previous researches,
is established by mathematical and explicit describing the equation of
ILCs. This model accurately reflects the characteristic behaviour of a
bottle-shaped strut, and is able to provide an overall view and an
adequate theoretical basis of internal stress distributions within a
Fig. 2. Calculation model for ILC equations.
482
Engineering Structures 167 (2018) 481–493
bottle-shaped strut. Splitting force formulae for the struts with different
load area ratios and height-to-width ratios are obtained individually
aiming to make some improvement in this topic. The proposed for-
mulae are critically validated using test and FEA results, and a com-
parison with Code [1,11,12], Sahoo’s method [4], and He’s method [6]
is also carried out, and the results will show which method is closer to
the test and FEA results.
3. Splitting force analytic model
3.1. ILC equation in one end of bottle-shaped struts
Previous researches [4–6,9] tell us to describe the compression
stresses in an elastic body subjected to a point load (or a distributed
load on a finite dimension) can be represented by isostatic lines. Sahoo
[4] first presented a mathematical description of these isostatic lines
using a differential equation. So similar to those eminent previous re-
isostatic lines of compression in bottle-shaped struts. According to
Saint-Venant’s principle, the stresses distribution in a bottle-shaped
strut is disturbed within a distance approximately equal to the strut
width. Therefore, under a concentrated load, plane b × b can be defined
as the dispersion zone. Fig. 2(a) shows the dispersion of compression in
the region of the struts under a concentrated load. Obviously, there are
infinite ILCs throughout the disturbed region of the struts, which con-
stituted a CDM under the concentrated load. To calculate the transverse
stresses, the CDM is defined as the mathematical model of principal
compressive-stress trajectories. Three assumptions on geometric and
physical boundary conditions of the ILCs of the CDM are given as fol-
lows.
Assumption 1. The ILCs are uniform distributed in both sections A-A′
and B-B′ as shown in Fig. 2(a). Based on geometric similarity theory, if
for an ILC with a vertical coordinate of yi at section B-B′, the vertical
coordinate at section A-A′ can be determined by times a/b (loading area
ratio), which can be written as
y|x = 0 = yi ·a/b,y|x = b = yi (2a)
Assumption 2. The ILCs must be perpendicular to the section B-B′,
which means the slope of the ILCs equal to zero, that is
dy
=0
dx x=b (2b)
A. Yuan et al. Engineering Structures 167 (2018) 481–493

In assumption 2, Sahoo et al. [4], He et al. [6] both think first order
derivative of the equation of ILC is equal to zero at x = 0, that is,

dy
=0
dx x=0 (4)

Fig. 4 shows the different models proposed by Guyou, Sahoo et al.,

He et al. In Guyou’s model, the single force P is uniformly distributed
over a height 2a′, symmetrical about the axis of the beam OZ. He did
not consider the effect of the load plate, and ignored the friction action
between the load plate and concrete. So these isostatics at x = 0 are
parallel to the applied force. But in Sahoo’s and He’s paper, the applied
force firstly transferred to the load plate, then dispersed into the con-
Fig. 3. Distribution of the transverse tensile stresses for various loading area
ratios a/b.
crete. As a matter of fact, the friction force between the load plate and
concrete exists objectively due to different Poisson’s ratio. If we add a
load plate into our model, we should take this effect into consideration.
Assumption 3. Seen from the Fig. 2(a), the transverse stresses have two Guyou’s assumption is not applicable here.
points, D and E, intersection with x axis. Point E is located at section B- Moreover, we focus on a point near the load plate as shown in
B′, which is the interface between D-region and B-region. So at section Fig. 2(a). If a small square element is isolated, the vertical shear stress
B-B′, the transverse stresses are equal to zero [4,6,13]. For this reason, on the left side face will be produced due to the friction between the
d 2y steel load plate and concrete. Considering equivalent law of shearing
=0 stress and the action of the two pairs of normal compression stresses on
dx 2 x=b (2c)
the vertical and horizontal faces, the eight stresses now act on the
element as shown in Fig. 2(b). Based on Mohr stress circle theory, these
stresses can be combined into two pairs of inclined compression
The location of point D is variable according to the different loading
stresses, known as principal stresses, as shown in Fig. 2(c). Obviously,
area ratios (a/b), which is verified by Guyou [3] and Gergely [14] and
the direction of principal stresses is not perpendicular to the load plate.
widely accepted by researchers [15–17]. According to Guyou’s test
According to above discussions, we think the first order derivative
results [3], the transverse stresses distribution is given as shown in
of the equation of ILC is not equal to zero at x = 0 and discard this
Fig. 3. It can be seen that as the loading area ratios increase, the dis-
assumption in this paper.
tance from the load plate to point D also increases. If the distance from
In assumption 3, Sahoo [4] assumes that second order derivative of
the dead plate to point D is defined to equal to k·b, based on these
the equation of ILC is equal to zero at x = 0, which is thought as a
previous researches, the specified values of k for different loading area
questionable assumption by He et al. [6], so He et al. discarded this
ratios are given in Table 1.
assumption and established a fourth-order polynomial equation for any
Because the transverse stresses are proportional to the curvature of
ILCs. In this paper, we agree with He’s assumption. Moreover, we take
the ILCs, we have
the effect of loading area ratios into consideration, and assume that at
d 2y the point of D (see Fig. 2(a)), x = k·b, stress varying from compression
= 0 (0 < k < 1) to tension, the second order derivative of the equation of ILC is equal to
dx 2 x = k·b (2d)
zero. From point O to D, the transverse stresses under load plate are
Using above five boundary conditions, a fourth-order polynomial compression, the second order derivative of the equation of ILCs should
equation can be obtained for any ILCs: (see the detailed derivation be greater than zero, and these ILCs must convex to the symmetrical
process in Ref. [19]) axis. And from point D to E, the transverse stresses become tension, the
second order derivative of the equation of ILCs should be less than zero,
(b−a) x x 3 2(1 + k ) 2 6k a
y=⎧ 2
× ⎡ 3− x + x + 2(1−3k )⎤ + ⎫ ·yi and these ILCs must concave to the symmetrical axis. The point D is the
⎨
⎩ (1−2k ) b ⎢
⎣b b2 b ⎥
⎦ b⎬⎭ (3) boundary between them. Therefore, this assumption captures the
characteristic of the ILCs, but previous studies ignored this important
where
boundary condition.
b = the width of a strut;
a = the width of a bearing plate; 3.3. Characteristics of the CDM
yi = the vertical coordinate of any ILCs at section B-B′.
Based on the relationship between the curvature of the ILCs and the
The equation of any ILCs can be obtained from the general Eq. (3) transverse stresses, the distribution of transverse stresses σT(x) along
by substituting the appropriate value for the vertical coordinate yi. the axis of a strut can be derived as (see the detailed derivation process
in Ref. [19])
3.2. Assumption comparison with previous research b /2 d 2y 3(b−a) P
σT (x ) = ∫0 dx 2
σ0 dyi = −
2(1−2k ) tb4
[x 2−b (k + 1) x + kb2]
(5)
Table 2 shows a comparison between different authors’ assumptions
for the equations of ILCs, including Sahoo’s assumption, He’s assump- where
tion, and assumption in this paper. It can be seen that the first as-
sumption is identical, but assumptions 2 and 3 are different. σ0= normal stresses in section B-B′, σ0 = −P/(bt);
P = applied load along the axis of a strut;
Table 1 t = thickness of a strut.
k value according to loading area ratios (a/b).
For different loading area ratios, corresponding coefficient k from
a/b 0.1 0.2 0.3 0.5 0.75 0.9
k 0.09 0.13 0.18 0.20 0.23 0.26 Table 1 is chosen, and the profiles are presented in Fig. 5.
The proposed Eq. (5) has some characteristics as follows: (1) at

483
A. Yuan et al. Engineering Structures 167 (2018) 481–493

Table 2
Assumption comparison included in different literature.
Assumption 1 Assumption 2 Assumption 3

Sahoo d2y d2y

y|x = 0 = yi ,y|x = b =
b
a (y − )
i
b−a
2
dy
dx x = 0
= 0,
dy
dx x = b
=0
dx 2 x = 0
= 0,
dx 2 x = b
=0

He a
y|x = b = yi ,y|x = 0 = yi · dy
= 0,
dy
=0 d2y
b dx x = 0 dx x = b =0
dx 2 x = b
This paper a
y|x = b = yi ,y|x = 0 = yi · dy
=0 d2y d2y
b dx x = b = 0, =0
dx 2 x = k·b dx 2 x = b
Comparison Identical Difference Difference

*To compare with each other, the same mean with the different symbols in different literature is unified using this paper’s symbol.

Z
B q1 E' C
E

b
6
q2

b
6
I 1 q3
2

b
6
P O 3 No equation
b

a
4

b
6
5
6
q4

b
6
q5
q6

b
6
A D
b
Lateral Stresses

Tension
No equation

Compression

(a) Guyou
b/2 0.27b 0.23b b/3 b/3 b/3
P/9 P/9
B C B C
P/9 P/9
P/9 Tb,1
P/9
P/9
P P/9
yi

P/9 P
yi
b

P/9
P/9
(2 yi − b )(b − a )x 3
(5b )
y = yi +
4ab 5
2
− 30bx + 12 x 2 (b − a )x 2 3 2 8 a
y = yi 3 2
x − x+6 +
A D b b b b
P/9
A D
Tension P/9
Lateral Stresses

x Tension
Lateral Stresses

x
15P a
σ T = - 4 1 − (b 2 x − 3bx 2 + 2 x 3 ) 3(b − a )P
2b t b σT (x) = − (3 x 2 − 4bx + b2 )
Compression Compression 2b 4t

(b) Sahoo (c) He and Liu

Fig. 4. Model comparison proposed by Guyou, Sahoo and He.

x = k·b, σT(x) = 0, which means the stresses varying from compression with the model proposed by Sahoo et al. [3] He et al. [6] (see Fig. 4). It
to tension; (2) at x = b, σT(x) = 0, which means stresses transition from can be seen that there are significant differences among them. Firstly, in
D-region to B-region; (3) at x = 0, the transverse stress is compression the model of Sahoo, the transverse stress below the load plate is equal
and reaches its maximum value; (4) when the bearing plate covers the to zero, while in He’s and this paper’s models, the transverse stress
entire width of the panel, a = b, σT(x) = 0, which means that the below the load plate reaches the maximum compression value, which
curvatures of all ILCs are equal to zero, and all the ILCs are straight was in a good agreement with the Guyon’s model shown in Fig. 2(a).
lines; and (5) with the increase of loading area ratios, the maximum Secondly, the turn point from compression to tension locates at x = b/2
transverse tensile stress decreases too, which is in a good agreement in the model of Sahoo and at x = b/3 in the model of He et al. While in
with the Guyon’s trend as shown in Fig. 3. the model of this paper, the turn point from compression to tension
The CDMs proposed in this paper are illustrated in Fig. 6. Compared locates at x = k·b, where the coefficient k varies according to the

484
A. Yuan et al. Engineering Structures 167 (2018) 481–493

0.4 where Tb is the splitting force in one end of a bottle-shaped strut

without D-region superposition.
0.3 The location of the splitting force db can be expressed as follows:
(see the detailed derivation process in Ref. [19])
0.2 b
∫k·b xσT tdx b (k + 1)
db = =
Stresses (MPa)

b
∫ σT tdx 2
0.1 k·b (7)
Eq. (7) further confirms that the location of the splitting force is
0.0 related to the loading area ratios.
0.0 0.2 0.4 0.6 0.8 1.0
-0.1 Distance from load plate 4. Transverse stresses and splitting force for Bottle-shaped struts
with different height-to-width ratios
-0.2
a/b=0.10, k=0.09 a/b=0.25, k=0.13
4.1. Force decomposition and stresses superposition
a/b=0.50, k=0.20 a/b=0.75, k=0.30
-0.3
This study investigated five types of bottle-shaped struts: (1) h/
Fig. 5. Transverse stress distribution for different loading area ratios. b ≥ 2, (2) 1 + k ≤ h/b < 2, and (3) 1 ≤ h/b < 1+k, (4) 2 k < h/
b < 1, (5) 0 < h/b ≤ 2 k. Based on the equivalent principle of force, a
bottle-shaped strut under an axial load can be decomposed into two
struts with one end subjected to a concentrated load, the other end
subjected to a uniform load and one strut with two ends subjected to a
uniform load. The typical force decomposition is shown in Fig. 7. Also,
the transverse stresses in a bottle-shaped strut can be treated as the
combination of three sub-models: two simple analytic models above
mentioned and one model under uniform pressure. In the latter sub-
model, there is no transverse stresses distribution. Therefore, using the
principle of superposition [18], the resultant stresses are equal to the
former two sub-models stresses superposition as shown in Fig. 8.

4.2. Transverse stresses and splitting force formula

(1) Height-to-width ratio (h/b) ≥ 2

If the height-to-width ratio (h/b) of a bottle-shaped strut is greater

than 2, there is a B-region zone between the two ends. Therefore, the
ILCs of the strut end have not affected each other. The transverse stress
can be expressed as Eq. (8),

⎧ σT (x ) 0<x⩽b
σT,1 (x ) = 0 b < x ⩽ h−b
⎨
⎩ σT (h−x ) h−b < x ⩽ h (8)
and the splitting force is also given by Eq. (9).
(k−1)3P
Tb,1 (x ) = Tb = (1−a/ b)
4(1−2k ) (9)
Fig. 6. CDMs for one end of a bottle-shaped strut. where
3(1−a/ b) P 2
σT (h−x ) = {x + [(k + 1) b−2h] + (h−b)(h−bk )}
different loading area ratios. There is no doubt that the model in this 2(1−2k ) tb3
paper is the closest to the Guyon’s model among these three methods.
Finally, the location of the splitting force in the model of Sahoo et al. The features of transverse stresses in this strut may attribute to
[4] and He et al. [6] is 0.79b and 0.67b away from the load plate, double peaks distribution and each end of the strut has a whole CDM,
respectively. However, the location of the splitting force in the model of which is shown in Fig. 9(a) and (b). There is no superposition region of
this paper is related to the loading area ratios. Once the loading area transverse stresses produced by each end of load. Therefore, the split-
ratio is chosen, the coefficient k is determined, and the location of the ting force at each end of the struts is equal to Tb.
splitting force in the model of this paper can be worked out.
(2) k + 1 ≤ Height-to-width ratio (h/b) < 2

3.4. Splitting force and its resultant location When the height-to-width ratio of a bottle-shaped strut is less than 2
and greater than k + 1, there is a superposition region of transverse
By integrating the transverse tensile stresses along the axis of the stresses produced by each end of the load. The transverse stresses can
struts, the total transverse force below the axis of the struts can be be expressed as Eq. (10)
determined by (see the detailed derivation process in Ref. [19])
⎧ σT (x ) 0 < x ⩽ h−b
b (k−1)3P σT,2 (x ) = σT (x ) + σT (h−x ) h−b < x ⩽ b
Tb = ∫k·b σT tdx =
4(1−2k )
(1−a/ b)
(6)
⎨
⎩ σT (h−x ) b<x⩽h (10)

485
A. Yuan et al. Engineering Structures 167 (2018) 481–493

Fig. 7. Force condition decomposition.

and the resultant splitting force is also derived as Eq. (11) 3(1−a/ b) P
σT (x ) + σT (h−x ) = [2x 2−2hx + 2kb2 + h2
h−b b 2(1−2k ) tb3
Tb,2 (x ) = ∫k·b σT (x ) tdx + ∫h−b (σT (x ) + σT (h−x )) tdx h−b
−bh (k + 1)] ∫k·b σT (x ) tdx
h − k·b P (1−a/ b)
+ ∫b σT (h−x ) tdx = 3
2b (1−2k )
{[b (1 + k )−h]2 [2h
=
P (2h−5b + bk )(b−h + bk )
(1−a/ b);
4b3 (1−2k )
+ b (k−5)] + (2b−h)2 [b (1 + 3k )−2h]} (11)
where

Fig. 8. Superposition of transverse tension.

486
A. Yuan et al. Engineering Structures 167 (2018) 481–493

Fig. 9. Transverse stresses distribution for different height-to-width ratios.

b P (2b−h)2 (b−2h + 3bk ) h− (2b−h)(h−2bk ) h+ (2b−h)(h−2bk )

∫h−b (σT (x ) + σT (h−x )) tdx = 2b3 (1−2k )
(1−a/ b); x3,1 = ;x3,2 =
(13)
2 2

h − k·b P (2h−5b + bk )(b−h + bk )2 The transverse stresses can be expressed as Eq. (14)
∫b σT (h−x ) tdx =
4b3 (1−2k )
(1−a/ b).
⎧ σT (h−x )−σT (x ) x3,1 < x ⩽ k ·b
The feature of transverse stresses is that as the height-to-width ratio σT,3 (x ) = σT (x ) + σT (h−x ) k ·b < x ⩽ h−k ·b
⎨
gets smaller and smaller from 2 to k + 1, the superposition stresses ⎩ σT (x )−σT (h−x ) h−k ·b < x ⩽ h−x3,2 (14)
distribution varies from three peaks distribution to one peak or almost
uniform distribution as shown in Fig. 9(c) and (d). However, even if the and the resultant splitting force can be written as Eq. (15)
transverse stresses distribution is different according to different height- k·b h − k·b

to-width ratios, the tensile stresses along the strut axis are super-
Tb,3 (x ) = ∫x 3,1
(σT (h−x )−σT (x )) tdx + ∫k·b (σT (x ) + σT (h−x )) tdx
position, and the resultant transverse force is a constant equal to 2 Tb. h − x3,2
+ ∫h−k·b (σT (x )−σT (h−x )) tdx
(3) 1 ≤ Height-to-width ratio (h/b) < k + 1 P (1−a/ b)
= 3 {3[b (k + 1)−h]2 (h−2bk ) + (h−2bk )2 [b (3 + k )−2h]}
2b (1−2k )
If the height-to-width ratio is greater than 1 and less than k + 1, a (15)
part of the tension stresses will be offset. The typical transverse stresses
where
distribution is shown in Fig. 9(e). To calculate the magnitude of the
resultant transverse force, we should know the initial zero point loca- 3(b−a) P
σT (x )−σT (h−x ) = {[2h−2b (k + 1)] x + bh (k + 1)−h2}
tion for integration, where tensile stress is equal to the compression 2(1−2k ) tb4
stress. It can be written as
k·b h − x3,3
σT (x ) = −σT (h−x ) (12) ∫x 3,1
(σT (h−x )−σT (x )) tdx = ∫h−k·b (σT (x )−σT (h−x )) tdx

Solution for the Eq. (12), the location of the initial zero point can be 3P (1−a/ b)
= [b (k + 1)−h]2 (2bk−h)
obtained 4b3 (1−2k )

487
A. Yuan et al.
h − k·b P (1−a/ b)
∫k·b (σT (x ) + σT (h−x )) tdx =
2b3 (1−2k )
(h−2bk )2 (2h−3b−bk )
The feature of transverse stresses is that the part of transverse ten-
sion stresses was offset by the part of transverse compression stresses
under the load plate. The resultant of the splitting force is a little
smaller than 2 Tb. The typical transverse stresses distribution is shown
in Fig. 9(e) and (f).
(4) 2 k ≤ Height-to-width ratio (h/b) < 1
When the height-to-width ratio (h/b) is less than 1, there is not
enough length to disperse the load force. Here, we use truncation skill
to deal with the stresses distribution, that is, if the height of the bottle-
shaped strut is less than the width of the struts, the stresses distribution
from the load plate to the distance equal to the height will be kept, and
the left will be cut off. Because the maximum value of the transverse
stresses is the compression stress under the load plate, the transverse
compression stresses under the load plate cannot be totally offset and
the transverse stresses under the load plate remain the compression
stresses. A little far away the load plate, the transverse compression
stresses decrease, and at the location where the transverse compression
stress is equal to the transverse tension stress, the transverse compres-
sion stresses become zero. The location can be determined by the
equation written as
σT (x ) = −σT (h−x )
Solution for the Eq. (16), the location where the transverse stress is
equal to the zero can be obtained
x 4,1 = x3,1 x 4,2 = x3,2
The transverse stresses can be expressed as Eq. (18)
⎧ σT (h−x )−σT (x ) x 4,1 < x ⩽ k ·b
σT,4 (x ) = σT (x ) + σT (h−x ) k ·b < x ⩽ h−k ·b
⎨
⎩ σT (x )−σT (h−x ) h−k ·b < x ⩽ h−x 4,2
and the splitting force can be written as Eq. (19)
k·b h − k·b
Tb,4 (x ) = ∫x 4,1
(σT (h−x )−σT (x )) tdx + ∫k·b
h − x 4,2
+ ∫h−k·b (σT (x )−σT (h−x )) tdx
P (1−a/ b)
= {3[b (k + 1)−h]2 (h−2bk ) + (h−2bk )2 [b (3 + k )−2h]}
2b3 (1−2k )
where the transverse stresses Eq. (18) and the splitting force Eq. (19)
are similar to the Eqs. (14) and (15) except the location of stress zero
x4,1, x4,2 and the value of k.
The feature of transverse stresses is that there is not a whole CDM in
the strut due to the height of the struts less than the width of the struts,
and the transverse stresses distribution has to be cut off the part which
exceeds the height of the struts. The resultant of the splitting force is
also less than 2 Tb. The typical transverse stresses distribution is shown
in Fig. 9(g).
(5) 0 < Height-to-width ratio (h/b) ≤ 2k
If the height-to-width ratio (h/b) is greater than zero and less than
2k, there are only the transverse compression stresses left along the
strut axis. The feature of transverse stresses is attributed to that the
stresses after superposition are also the compression stresses, and there
is no splitting force produced. The typical transverse stresses distribu-
tion is shown in Fig. 9(h).
Engineering Structures 167 (2018) 481–493
4.3. Simplify equations for practical application
Above proposed Eqs. (11), (15) and (19) are very complex. So these
equations have to be simplified for practical application. According to
the principle of superposition, the lower the bottle shaped strut be-
comes, the smaller the splitting force will be. So for five cases, we
substitute the upper boundary into each equation, then get the max-
imum value of the splitting force for each case to control the design.
The simplified equations are listed in Table 4.
4.4. Examination for EN 1992-1-1: 2004
Above five scenarios discuss the stresses distribution and the re-
sultant of the splitting force. It can be found that if k + 1 ≤ Height-to-
width ratio (h/b) < 2, no tensile stresses along the strut axis are offset,
and the resultant transverse force is a constant equal to 2 Tb. However,
if using Eq. (1) (Article 6.5.3, Eq. (6.69) in EN 1992-1-1:2004) to cal-
culate the tensile force, it will be decreased with the lower height of the
strut. Meanwhile, when 0 < Height-to-width ratio (h/b) ≤ 2 k, there is
no splitting force produced. It is recommended that these two scenarios
be considerated.
5. Other theoretical equations
Current several codes [1,11,12], Sahoo et al. [4] and He et al. [6] all
(16) give the equations of splitting force and its location as shown in Table 3.
If we make a comparison among different methods above men-
tioned for the calculation of transverse tension force. Firstly, there are
different professional terms in different components. In the anchorage
zone, the transverse tensile force is called the bursting force, while in
(17) the bottle-shaped struts, it is named the splitting force. Here, we follow
the habit of this usage. Secondly, it can be seen that these formulae of
splitting force in Table 3 all consider the effect of the loading area ratio
(a/b), but none of them further consider the effect of the different lo-
cation where the stresses turn compression into tension due to the
different loading area ratios. Finally, for each formula, there is a con-
(18) stant location of the resultant transverse tension force, regardless of the
varied loading area ratios. In fact, the location of the resultant splitting
force has a close relationship with the loading area ratios as shown in
Eq. (7).
(σT (x ) + σT (h−x )) tdx
6. Validation with experimental and FEM results
6.1. FEM validation
In order to check the accuracy of the equations proposed by Sahoo
(19)
et al. [4], He et al. [6] and authors, several typical bottle-shaped struts
were analysed using finite element method and their proposed equa-
tions. Fig. 10 shows two-dimensional finite element models for the
anchorage zone with different h/b. The boundary at the far-end is re-
strained to evaluate Eq. (5) as shown in Fig. 10(a). And several self-
balance finite element models with different height-to-width ratios are
built to evaluate the accuracy of Eqs. (9), (11), (15) and (19) as shown
in Fig. 10(b), respectively. In these finite element models, the concrete
and load plate are modelled by four-node plane stress elements. The
young’s modulus and Poisson’s ratio of concrete are 3.0 × 104 MPa
[4.35 × 103 ksi ] and 0.2, and these of steel are 2.0 × 105 MPa
Table 3
Formulae for the splitting force and its location.
ACI/AASHTO/CEB FIP Sahoo He and Liu
Splitting force Tb = 0.25P (1−a/ b)
Tb =
15P
64 (1− )
a
b
Tb =
2P
9 (1− )
a
b
Location db = 0.5b db = 0.79b db = 0.67b
488
A. Yuan et al. Engineering Structures 167 (2018) 481–493

Table 4
Simplified equations for practical application.
Cases h/b ≥ 2 k + 1 ≤ h/b < 2 1 ≤ h/b < k+1 2 k < h/b < 1 0 < h/b ≤ 2 k

Equations Eq. (9) Eq. (11) Eq. (15) Eq. (19) 0

Assumption – h = 2b h = (1 + k)b h=b –
Simplified equation (k − 1)3P (k − 1)3P (k − 1)3P (1 − k + k2) P 0
(1−a/ b) (1−a/ b) (1−a/ b) (1−a/ b)
4(1 − 2k ) 2(1 − 2k ) 2(1 − 2k ) 2

Symbol “–” means no assumption.

[2.9 × 104 ksi] and 0.3. The concentrated load is applied as a uniform
load within the footprint of the load plate.
Fig. 11 compares the distributions of the transverse stresses ob-
tained along the strut axis for the results of the FEA, equations of Sahoo
et al. [4], He et al. [6], and Eq. (5) in this paper, with the load area
ratios of 0.2, 0.3, 0.5 and 0.75. These diagrams indicate that there is a
substantial discrepancy among the transverse stresses given by the re-
sults of FEA and different equations. It can be seen that Eq. (5) agrees
better with the FEA results than the previous analytical equations
proposed by Sahoo et al. [4] and He et al. [6]. Especially, Eq. (5) can
not only reflect the variation of the turn point from the compression to
the tension according to the different loading area ratios, but also have
a good agreement with FEA results while other equations cannot realize
this point. But it also can be seen that there is a big difference to de-
scribe the transverse compression stresses distribution. It may attribute
to that a simple polynomial equation cannot wholly consider the dis-
tribution of compression stresses and tension stresses at the same time.
Considering the bottle-shaped struts with different loading area
ratios, Table 5 presents the splitting forces and their locations calcu-
lated by the equations of Sahoo et al. [4], He et al. [6], Eq. (5) and the
results of FEA. It can be seen that when the a/b value is equal to 0.3, 0.4
and 0.5, the splitting force predicted by Eq. (5) with the Tb/P values of
0.151, 0.107 and 0.129 is the closest to the results of FEA with the Tb/P
values of 0.151, 0.107 and 0.131, respectively. Furthermore, the loca-
tions of the resultant of splitting forces, calculated by Eq. (7) proposed
in this paper, are also in good agreements with the FEA analysis results.
When loading area ratios change, the locations of the resultant of
splitting forces also vary accordingly. Only Eq. (7) can reflect such
phenomenon while others have a constant value. The maximum re-
lative error is 52% and 29% for the equations of Sahoo et al. and He
et al.
To extensively evaluate the accuracy of the proposed formulae, Eqs.
(9), (11) and (15), based on FEA results, Table 6 give the comparison
between FEA results and the value calculated by above equations for
bottle-shaped struts with different a/b (0.2, 0.4, 0.5) and different h/b
(1.0, 1.5, 2.0). Seen from the Table 6, the ratios of Tb /FEA are in a
range of 0.957–1.195. It indicates the proposed Eqs. (9), (11) and (15)
have a good precision, which can predict the magnitude of splitting
force exactly.
6.2. Experimental validation
(1) Test specimens
In this paper, to validate the superposition effect and splitting force
of the bottle-shaped struts, 6 thin concrete panels, measuring 900 mm
[35.43 in.] high, 900 mm [35.43 in.] width, and 150 mm [5.91 in.]
thick, with different loading area ratios and different reinforcement
angles, were tested to failure. The height-to-width ratio of the bottle-
shaped struts was equal to 1, therefore Eq. (15) will be critically
checked. The different loading area ratios had three levels, 0.27, 0.2
and 0.17, that was, the load plate length was equal to 240 mm [9.45
in.], 180 mm [7.09 in.] and 150 mm [5.91 in.], respectively. The dif-
ferent reinforcement angles had two kinds of degree, 0° and 45°. The
details of test panels are shown in Fig. 12. The steel reinforcement was
Grade HPB300 with yield and ultimate strengths of 300 MPa [43.51
ksi.] and 420 MPa [60.91 ksi.], respectively. And the tested cube con-
crete compressive strength was, on average, 49.59 MPa [7.19 ksi.] with
a corresponding standard deviation of 0.64 MPa [0.09 ksi.].
(2) Test setup and instrument
Using a 320 t [705 klb.] hydraulic jack and a self-equilibrating
loading frame with maximum capacity 4000 kN [899.2 kip], the panels
were tested to failure under axial compressive load. The loads were
applied through a pair of 50 mm [1.97 in.] thick steel-bearing plates
with different lengths and placed at the top and bottom edges. The
transverse strains along the longitudinal axis of the panel were mea-
sured by the concrete strain gauge. The magnitude of the load was
recorded by the load cell. The whole test setup and instrument layout
are shown in Fig. 13(a) and (b).
(3) Test results
To calculate the measured transverse tensile stresses (σy) of the
struts, 2D Hooker’s law is used
Ec
σy = (εy + νεx )
1−v 2 (20)

(a) For evaluate Eq. (5) (b)For evaluate Eqs. (9), (11) and (15)
Fig. 10. Finite element model for bottle-shaped struts.

489
A. Yuan et al. Engineering Structures 167 (2018) 481–493

1.0
0.5 0.5

0.0 0.0
-0.5
-0.5
-1.0
-1.5 a/b=0.2 -1.0 a/b=0.3

σT/σ0
σT/σ0

σ0=P/(ht) σ0=P/(ht)
-2.0 Eq. (5) Eq. (5)
-1.5
-2.5 FEA FEA
-3.0 P
x Sahoo et al. -2.0 P
x Sahoo et al.
s0 s0

h
a
h

He et al.
a
-3.5 He et al.
-2.5
-4.0 l=h
l=h
-4.5 -3.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Relative distance, x/h Relative distance, x/h

(a) a/b = 0.2 (b) a/b = 0.3
0.50 0.25

0.25

0.00 0.00

-0.25

a/b=0.75
σT/σ0

-0.50 -0.25
σ0=P/(ht) a/b=0.5 σ0=P/(ht) Eq. (5)
-0.75 Eq. (5) σT/σ0 FEA
x FEA x Sahoo et al.
-1.00 P s0 -0.50 P
Sahoo et al. s0
h
a

He et al.

h
a
-1.25 He et al.
l=h l=h
-1.50 -0.75
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Relative distance, x/h Relative distance, x/h
(c) a/b = 0.5 (d) a/b = 0.75
Fig. 11. Transverse stress distributions in a bottle shaped strut with different loading area ratios.

where
εy = the transverse tensile strain of the struts;
εx = the longitudinal compression strain of the struts;
ν = the concrete Poisson’s ratio, equal to 0.2;
Ec = the elastic modulus, which can be taken as
Ec = 4730 fc′ (MPa
The transverse stresses along the longitudinal axis of the struts are
shown in Fig. 14. It can be seen that under the load plate, it is a com-
pression zone, and a little far away from the load plate, it becomes a
tension zone. The distribution of the theoretical transverse stresses
shows a good agreement with the FEA and test results. According to
Sahoo’s method, the maximum tensile stress does not exceed the
splitting tensile strength of concrete, ft . So assume if calculated stress
σy > ft , then σy = ft . Here, ft = 0.56 fc′ (MPa), fc′ = the cylinder
compressive strength of concrete (1 MPa = 0.145 ksi). Then the total
test resultant transverse force can be calculated by
n
Ttest = ∑ σyi li t
i=1 (22)
where li = the length of the rectangle area with the stresses equal to σyi.
Eq. (15) is used to calculate the theoretical resultant transverse
force. when h = b = 900 mm [35.43 in.], a = 150 mm [5.91 in.],
180 mm [7.09 in.] and 240 mm [9.45 in.], we can calculate the value of
(21) k from Table 1, that is, k = 0.118, 0.13 and 0.165, respectively.
The proposed method, Code provisions, Sahoo’s method and He’s
method are used to calculate the splitting force of the 6 bottle-shaped
struts. The struts considered in this study have loading area ratios
ranging from 0.17 to 0.27. A summary of these struts is presented in
Table 7. It can be seen that compared with the test results, the proposed
method and He’s method have good agreements with mean values of
0.99, 0.99 and standard deviation values of 0.02, 0.03, respectively.
While code provisions and Sahoo’s method show a little conservative
with mean values of 1.12, 1.05 and standard deviation values of 0.04,
0.04, respectively.
As far as the distances of turn point from the load plate are con-
cerned, the average measured results, FEA results and theoretical

Table 5
Splitting force and their locations a bottle-shaped strut with different a/b.
a/b = 0.3 a/b = 0.4 a/b = 0.5

Tb/ p Location db/ h Tb/ p Location db/ h Tb/ p Location db/ h

FEA 0.151 0.52 0.107 0.54 0.131 0.56

Eq. (5) 0.151 0.59 0.107 0.595 0.129 0.6
Sahoo et al. 0.164 0.79 0.117 0.79 0.141 0.79
He et al. 0.156 0.67 0.111 0.67 0.133 0.67

490
A. Yuan et al. Engineering Structures 167 (2018) 481–493

Table 6
Splitting force for bottle-shaped struts with different a/b and h/b.
a/b = 0.2 a/b = 0.4 a/b = 0.75

Tb FEA Tb/FEA Tb FEA Tb/FEA Tb FEA Tb/FEA

kN kN kN kN kN kN

h/b = 1, Eq. (15) 567.6 517.6 1.097 812.2 688.7 1.179 617.1 516.5 1.195
h/b = 1.5, Eq. (11) 569.5 573.9 0.992 822.9 822.8 1.000 634.1 655.9 0.967
h/b = 2, Eq. (9) 569.5 575.6 0.989 822.9 827.0 0.995 634.1 662.2 0.957

(1 kN = 0.2248 kips).

Section 6.1.

7. Summary and conclusions

In this paper, a quantitative compression-dispersion model is es-

tablished for a bottle-shaped strut with different loading area ratios.
Then considering different height-to-width ratios, individually devel-
oped compression-dispersion models are established to calculate the
splitting force of the struts. Finally, the proposed formulae to calculate
splitting force in bottle-shaped struts are verified. The following con-
clusions are drawn from above analysis results.
. (1) A new ILCs equation is theoretically and mathematically obtained,
(a) = 0o (b) = 45o which considers two important geometric and physical characters
Fig. 12. Test specimens. of the isostatic lines of compression: the first order derivative of the
equation of ILC is not equal to zero at x = 0 and the second order
derivative of the equation of ILC is equal to zero at x = k·b.
results predicted by Eq. (14) are listed in Table 8. Seen from Table 8, it
Established CDMs for bottle-shaped struts are in good agreements
can be found that the results obtained from the finite element analysis
with Guyon’s CDM that with the increase of loading area ratios, the
are closest to the test results. The maximum error is only 10% for a/
maximum transverse tensile stress decrease and the turn point from
b = 0.17. Guyou’s test results are greater than the FEA and test results
the compression stresses to the tension stresses changes corre-
because Guyou’s results are obtained from a strut with one end under
spondingly;
concentrated load, the other end under uniform load, not considering
(2) Based on the force decomposition and stresses superposition, the-
the effect of the superposition effect. The results obtained from Eq. (14)
oretical equations are established to calculate splitting force of the
are smaller than the FEA and test results. It may be caused by a big
struts with different height-to-width ratios. Simplified equations for
difference in describing the transverse compression stresses distribution
different scenarios are given and a recommendation for EN 1992 is
between the theoretical Eq. (5) and the FEA results, as mentioned in

(a) Instrument layout (b) Test setup

Fig. 13. Test instrument layout and setup.

491
A. Yuan et al. Engineering Structures 167 (2018) 481–493

(a) a/b = 0.17 (b) a/b= 0.2

(c) a/b = 0.27

Fig. 14. Distribution of transverse stresses in a bottle shaped strut with different loading.

Table 7 than that obtained from test and FEA results.

A comparison of different method with test results. (4) The proposed Eqs. (5), (9), (11) and (15) are carefully verified using
Specimens a/b Pcr, kN Ttest, kN Tmethods/Ttest
test results and FEM analysis results. It is found that the proposed
methods show closer to the test and FEA results than the other
Proposed Code Sahoo He methods.

BSS1-150 0.17 600 234 0.95 1.06 1.00 0.95

BSS1-180 0.20 680 248 0.97 1.10 1.03 0.97
BSS1-240 0.27 800 254 0.99 1.15 1.08 1.02 8. Future research
BSS2-150 0.17 750 279 1.00 1.12 1.05 0.99
BSS2-180 0.20 795 285 0.99 1.12 1.05 0.99 The design of the bottle-shaped struts is complex, and efforts using
BSS2-240 0.27 802 250 1.01 1.17 1.10 1.04
CDM to calculate the splitting force will continue. It will be desirable to
Average 0.99 1.12 1.05 0.99
Standard deviation 0.02 0.04 0.04 0.03 take the friction between the load bearing plate and the concrete into
consideration, further improve the CDM and reveal the mechanism of
Ttest lists the results calculated by a trapezoid model. (1 kN = 0.2248 kips). the force transfer. Additionally, it is common for a girder with multiple
anchorages, and an attempt for the application of the described equa-
Table 8 tions of the presented approach for the case of two adjacent anchorages,
Distance of turn point from the load plate (units: mm). symmetrically or asymmetrically placed at the girder end with various
a/b = 0.17 a/b = 0.20 a/b = 0.27 dimensions, would really be of great interest. Finally, it is re-
commended that a piecewise equation or other reasonable equations
Eq. (14) 57 63 82 instead of a polynomial equation be used to describe the transverse
Test results 77 83 89
stresses distribution in the future research.
FEA results 69 76 89.5
Guyou’s results 106 117 148.5

given;
(3) Eq. (5) can not only reflect the variation of the turn point from the
compression to the tension according to the different loading area
ratios, but also have a good agreement with FEA results. Con-
sidering the superposition effect, the location of the turn point
calculated by theoretical equations will be closer to the load plate
Acknowledgments
This work in this paper is supported by the Province Natural Science
Foundation of Jiangsu (Grant No. BK2011748) and Science and
Technology Planning Project of Department of Transportation of
Jiangsu Province (Grant No. 2013Y11). The authors would like to thank
the technical reviewers and editors for their constructive comments.

492
A. Yuan et al.
References
[1] ACI Committee 318. Building Code Requirements for Structural Concrete (ACI 318-
14) and Commentary (318R-14). American Concrete Institute, Farminton Hills,
Mich., 2014, pp. 520.
[2] Biondini F, Bontempi F, Malerba PG. Stress path adapted strut and tie models in
cracked and uncracked RC elements. In: 1st International conference on advances in
structural engineering and mechanics, Seoul, South Korea, August 23–25; 1999. p.
955–62.
[3] Guyon Y. Prestressed concrete. London: Contractor’s Record; 1953.
[4] Sahoo DK, Singh B, Bhargava P. Investigation of dispersion of compression in bottle
- shaped struts. ACI Struct J 2009;106(2):178–86.
[5] Windisch A. Discussion of ‘investigation of dispersion of compression in bottle-
shaped struts’ by Dipak Kumar Sahoo, Bhupinder Singh, and Pradeep Bhargava. ACI
Struct J 2010;107(1):124–5.
[6] He ZQ, Liu Z. Investigation of splitting forces in anchorage zones: compression-
dispersion models and unified design equation. J Bridge Eng 2011;16(6):820–7.
[7] Breen JE, Burdet O, Roberts C, Wollmann G. Anchorage zone reinforcement for
post-tensioned concrete girders. Final Rep. For NCHRP 10–29. Austin (TX): Univ. Of
Texas at Austin; 1994.
[8] Brown MD, Bayrak O. Minimum transverse reinforcement for bottle-shaped struts.
ACI Struct J 2006;103(6):813–21.
[9] Zhou LY, Liu Z, He ZQ. Further investigation of transverse stresses and bursting
493
Engineering Structures 167 (2018) 481–493
forces in post - tensioned anchorage zones. Struct Concr 2015;16(1):84–92.
[10] Ghanei AH, Aghayari R. Experimental investigation on effect of aspect ratio on
behavior of isolated reinforced concrete struts. ACI Struct J 2016;113(4):643–54.
[11] Ceb FIP. Model Code 1990. London: Thomas Telford Services Ltd; 1993.
[12] American Association of State Highway and Transportation Officials (AASHTO).
AASHTO LRFD Bridge Design Specifications, Section 5: Concrete Structures, 6th
Edition and 2013 interims, Washington, D.C.; 2013. p. 262.
[13] Schlaich J, Schȁfer K, Jennewein M. Toward a consistent design of structural con-
crete. PCI J 1987;32(3):74–150.
[14] Gergely P, Sozen MA. Design of anchorage zone reinforcement in prestressed con-
crete beams. PCI J 1967;12(2):63–75.
[15] Wollmann GP. Anchorage zone in post-tensiloned concrete structure Ph.D Thesis
Austin: The University of Texas; 1992.
[16] Iyengar K, Yoganada C. A three-dimensional stress distribution problem in the
anchorage zone of a post-tensioned concrete beam. Mag Concr Res
1966;18(55):75–84.
[17] Burdet O. Analysis and design of anchorage zones for post-tensioned concrete
bridges Ph.D. Thesis Austin (TX): University of Texas at Austin; 1990.
[18] He ZQ, Liu Z, Ma ZG. Explicit solutions of transverse tensions in deep beams: load-
path model and superposition principle. ACI Struct J 2014;111(3):583–94.
[19] Yuan Aimin, Xu Donghui, Qian Shoulong, Dai Hang. Derivation of the equation of
Isostatic line of compression and splitting force in a bottle-shaped strut. Data in
Brief; 2018 [in press].