Engineering Structures 221 (2020) 111087

Contents lists available at ScienceDirect

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

Discrete analytical model for lateral mechanical behavior of cable-saddle T

system in suspension bridges
Lu Wanga,b, Ruili Shenb, Niujing Maa,c, Songhan Zhangd, Lixiong Gua, , Ronghui Wanga
⁎

a
School of Civil Engineering and Transportation, South China University of Technology, 510641, China
b
School of Civil Engineering, Southwest Jiaotong University, 610031, China
c
State Key Laboratory of Traction Power, Southwest Jiaotong University, 610031, China
d
Department of Civil Engineering, Sichuan University, 610031, China

ARTICLE INFO

Keywords: Abstract: Frictional resistance, typically provided by the bottom and sides of saddle, is essential for counter-
Multi-tower suspension bridge poising unbalanced cable tension between adjacent spans. In most practices, however, the friction from the
Analytical model saddle sides is ignored due to the lack of an effective calculation method of lateral forces, giving rise to serious
Discrete cable wires anti-slip issues, especially for multi-tower suspension bridges. For this reason, the lateral forces between the
Lateral force
main cable and saddle are studied in this paper. A similar formula of lateral force in specification is first in-
Janssen formula
Frictional resistance
vestigated to reveal its theoretical basis and limitations. Considering each cable wire as an individual discrete
body, an analytical model with recursive algorithm is then developed, which allows to calculate the contact
forces layer by layer. Based on this, the effects of possible errors are quantified, and the sensitivity of lateral
forces to key parameters, such as friction coefficient, extra pressure, arrangement and diameter of the wires, is
subsequently discussed in detail. The results show that the existing formula evolved from the Janssen theory is
insufficient for anti-slip assessment of the main cable. The transverse gap of saddle trough has a limited increase
effect on the lateral force. As the wire arrangement widens, the lateral force grows rapidly and then approaches a
certain value. The lateral force is negatively related with the friction coefficient, and the extra pressure can
create great yet regular influence on the distribution of lateral pressure. Finally, the Taizhou Yangtze River
Bridge is taken as a case example for an application of the proposed model. It is found that 32.2% of the total
friction comes from the lateral friction when two vertical friction plates are used, confirming that the anti-slip
capacity of cable-saddle system can be greatly improved by adequately utilizing the lateral friction.

1. Introduction frictional resistance between the main cable and saddle. Unfortunately,
the friction provided by the saddle is usually too weak to meet the anti-
Due to their notable spanning capacity and an avoidance of costly slip requirements, especially when evaluated by the traditional sim-
shared anchorage, multi-tower suspension bridges have been re- plified method that ignores the lateral friction due to the absence of a
cognized as one of the most competitive solutions for crossing broad calculation method [8–11]. Ensuring sufficient and well-calculated
waters, and thus are gaining more and more attention [1–4]. However, frictional resistance is therefore critical to the design of multi-tower
in practice, the problems related to the anti-slip safety between the suspension bridges.
main cable and saddle excluded the extensive introduction of this As depicted in Fig. 1(b), due to the longitudinal bending transition,
bridge type [5–7]. the main cable with huge tensile force will induce great radial force
In a typical multi-tower suspension bridge (Fig. 1), the main cable towards the bottom of the saddle trough. Meanwhile, because the effect
continuously passes through all spans, with the saddles serving as in- of lateral restraint of saddle, the main cable under the radial force will
termediate constraints in the longitudinal direction. When live load is also create lateral force on the sides of the saddle trough. According to
concentrated on a single main span, the greatest tension difference can Coulomb’s friction law [12], making clear of the two normal forces is
be produced in the main cable between the two sides of the saddle. the premise for calculating the frictional resistance. For the radial force,
Since the main cable is merely placed in the saddle fixed atop the tower there has been an explicit formula, which is the ratio of cable tension to
[Fig. 1(b)], such tension difference can only be balanced by the saddle radius [10,13]. However, the knowledge in terms of the lateral

⁎
Corresponding author.
E-mail address: [email protected] (L. Gu).

https://doi.org/10.1016/j.engstruct.2020.111087
Received 24 December 2019; Received in revised form 1 July 2020; Accepted 4 July 2020
Available online 20 July 2020
0141-0296/ © 2020 Elsevier Ltd. All rights reserved.
L. Wang, et al. Engineering Structures 221 (2020) 111087

Fig. 1. Illustration of the investigated structures: (a) typical bridge layout and unbalanced loading condition; (b) longitudinal and cross-sectional views of cable-
saddle system and the concerned forces.

force is still limited. are subsequently performed for revealing the effects of some key factors
Hasegawa et al. [9] derived a formula for the nominal friction on the lateral force, involving diameter and arrangement of the wires,
coefficient of a combined saddle with a horizontal friction plate. friction coefficient and extra pressure. A typical multi-tower suspension
However, the effect of lateral friction was neglected due to the absence bridge is finally taken as a case example, and the frictional mechanism
of a calculation method for the lateral force. A seeming formula of and the effects of lateral friction are thus analyzed using the proposed
lateral pressure was found in a design specification [13], but this for- model.
mula is specifically used to check the lateral strength safety of the
saddles. Moreover, neither the theoretical basis nor the applicable 2. Investigation of the existing formula
scope of this formula was given, which puzzles the engineers in appli-
cation, and further clarification is needed. Zhang et al. [7,14] and The design specification for highway suspension bridge [13] gives a
Cheng et al. [3] successively proposed an analytical model and a nu- formula of lateral pressure for designing the saddle, without any ex-
merical method for evaluating the cable-saddle frictional resistance. In planation about its theoretical basis and applicable border. In this
the model or method, the main cable was assumed as an integral body, section, this formula (named existing formula hereafter) is theoretically
and the calculation of the lateral stress between cable and saddle was derived and clarified, with referring to the Janssen theory in granular
simplified as a plane strain problem. For double-cable multispan sus- mechanics.
pension bridges, Zhang et al. [15] presents an innovative saddle to
accommodate two main cables and analyzes the frictional resistance of
2.1. Description of the Janssen Theory
the saddle. Based on the bundle characteristic of the main cable, Wang
et al. [16] performed model tests for investigating the slip behavior
Silos are common structures for storing various granular materials
between cable and saddle. The test results indicated that the lateral
such as cement, sand, and cereal grains (Fig. 2). When the granules are
friction has a noticeable enhancement effect on the frictional resistance.
filled into the silo, the own-weight of granules and the possible extra
In order to analyze the slip behavior of layered cable wires in-
pressure will induce lateral pressure along the silo wall, which com-
corporating with horizontal friction plates, a discrete-body model was
plicates the design of silo wall. The analytical derivations of this issue
developed, allowing to present the contact forces between wires, and
were pioneered by Janssen [18], and the well-known Janssen formula
the layer-by-layer slip was confirmed by comparing to a model test
was developed based on the continuum medium model
[17]. This discrete-body model, which has been proved to be feasible in
the specific case, will be of greater significance if it is extended to be g µkh µkh
fy(h) = 1 e + k 0e
general. µ (1)
From the above, it can be concluded that the anti-slip issue of sus-
pension bridges is promising to be addressed by adequately utilizing the
lateral friction between the main cable and saddle, which is inspiring to
the bridge designers. Nevertheless, the distribution of lateral forces,
being quite important to the efficiency of the anti-slip approach, should
be precisely evaluated taking into account the interaction between
wires, prior to further practical applications. A general analytical
model, consisting of discrete bodies for describing the wires, is needed
for studying the lateral forces in broader cases. This remains so far an
open question to be solved, which is the motivation of the present work.
Therefore, this study aims at presenting an efficient analytical
model for analyzing the lateral force between the main cable and
saddle. In this study, the theoretical basis and limitations of a similar
formula in specification are first clarified, and an analytical model,
considering the discrete contact nature of cable-saddle system, is de- Fig. 2. Physical features and mechanical relationships between silo and gran-
veloped and then confirmed by error estimations. Sensitivity analyses ules.

2
L. Wang, et al. Engineering Structures 221 (2020) 111087

Fig. 3. Equilibrium analysis for infinitesimal cable element on the saddle.

where fy(h) = lateral pressure at depth h; = density; g = gravita- exemplify the derivation below. Being consistent with the homo-
tional acceleration; = hydraulic radius of the net horizontal area; µ genization strategy adopted by Janssen theory, the cable wires are as-
= friction coefficient between storage material and silo; k = ratio of sumed as a continuous medium with uniformly stressed. Then, the
the lateral pressure to the vertical pressure; and 0 = a uniform pres- gravity density ( g ) in Eq. (1) can be equivalent to the unit radial vo-
sure surcharge at the top. lume force ( fv ) of the cable. Recalling Eq. (3), fv can be expressed as
The following three assumptions are adopted in the Janssen theory: Nwc Tc
fv =
Nw bHc R (4)
1. The stresses are uniform over each horizontal cross section of the
granules, and the ratio k is a constant. where Tc = design value of cable tension; Nwc = number of cable wires
2. The granules are incompressible, and the silo is infinitely deep. in the center trough; Nw = total number of cable wires in the saddle; b
3. The frictional shear stress between granules and silo wall satisfies and Hc = width and height of the cable wires in the center trough,
the Coulomb’s failure criterion, i.e., the granules are at the onset of respectively.
sliding at any point of the wall [19]. Furthermore, according to the rectangular section of the saddle
trough (Fig. 4), it can be acquired that the cross-sectional area A = bHc ,
and the perimeter P = 2Hc + b . Thus, the hydraulic radius is given by
2.2. Derivation of the existing formula
A/ P [20], that is
The existing formula is to be derived based on the Janssen theory A bHc b
= = =
and the structural characteristics of cable-saddle system. As illustrated P 2Hc + b 2 + b/Hc (5)
in Fig. 3, a cable passes through a saddle, and the cable tensions on the
Since b Hc , Eq. (5) can be simplified to = b/2 .
loose and tight sides are denoted by Tcl and Tct , respectively. The free-
Follow the Assumption 1 of Janssen theory, the k-value could be
body diagram of an infinitesimal element d of the cable is plotted, for
interpreted as, for instance, the elastic redirection coefficient [21–23]
which the equilibrium of the element in the radial direction is expressed
as k=
1 (6)
d d
fz R·d = (Ts + dTs )sin + Tssin where = Poisson ratio. If = 0.25 is used for structural steels, then
2 2 (2)
k = 1/3 can be obtained.
where fz = unit radial pressure between the cable and saddle; R = Thus, substituting g = fv , = b/2 , and k = 1/3 into Eq. (3) gives
radius of saddle trough; and Ts = tension of the cable element. the general expression of lateral pressure between the cable and saddle
By neglecting the second-order term of Eq. (2), fz can be obtained as
fv b 2µh 1 2µh
T fy(Exih) = 1 e 3b + 0e 3b
fz = s 2µ 3 (7)
R (3)
where the upper index “Exi” represents throughout the paper that some
On the other hand, Fig. 4 illustrates the cross-section of a conventional quantities belong to the existing formula. In general, there is no extra
saddle structure. The saddle is divided into several troughs by spacers, pressure on the top, or the applied pressure is negligible compared to
and the cable wires are arranged in each trough regularly. The width of the cable self-pressure. Therefore, by removing the last item of Eq. (7),
each trough is normally equal, and the number of cable wires gradually the existing formula is finally obtained
decreases from the center trough to the side trough. Therefore, the only
difference between these troughs is in the height dimension. The center fv b 2µh
fy(Exih) = 1 e 3b
trough, containing the most wire layers, can thus be selected to 2µ (8)
Subsequently, the corresponding total lateral force can be determined
by integrating fy(h) along the height Hc
2µHc
Hc fv bHc 3fv b2 (1 e 3b )
FyExi = f y(Exih) dh =
0 2µ 4µ2 (9)

2.3. Discussions of the derivation

The above derivations reveal that the existing formula is evolved

from the Janssen formula, yet in terms of its rationality, some funda-
mental shortcomings can be found. First, it is well accepted that the
Janssen formula can give good prediction of pressures in the deep
Fig. 4. Cross-section of saddle structure and homogenization of cable wires. containers [24,22]. However, there is no deep or shallow concept for

3
L. Wang, et al. Engineering Structures 221 (2020) 111087

the saddle structures in the design specification [13], which makes the where C1 = a coefficient refined from the transfer rules of internal
applicability of the existing formula ambiguous. forces; µ w = friction coefficient between wires; = stacking angle
Besides, it is well known that k is the most critical parameter in the of wires, for the presented regular stacking pattern = 30°; fi,Lj and
Janssen formula [25,26]. However, prolonged disagreements still exist fiR, j = normal force transferred from the wire in the ith layer and jth
on the determination of its value [25,19]. Note that when deriving the column, namely the (i, j) wire in this study, to its lower-left (i + 1, j)
existing formula, the cable was assumed as a continuum and the wire and the lower-right (i + 1, j + 1) wire, respectively; f1,evj =
Poisson effect was introduced to interpret the k-value. Unfortunately, external vertical force acting on the top wire in the jth column; and
these simplifications obscured the essential divergence between the fi,wvj = self-generated vertical force of the (i , j ) wire, which can be
granules and the cable wires (Figs. 2 and 4). For the granules, due to calculated by referring to Eq. (3)
their intensive quantity, small size, random shape and ultralight self-
weight, the internal stacking pattern and force transmission are quite Tiw, j Tiw
,j
fi,wv
j = =
complicated yet negligible. The granules can thus be homogenized and Ri R+H hi (12)
assumed to act like a continuum at a macroscopic level [24]. In con- where Tiw, j
= tension of the (i, j) wire; Ri = radius of the ith layer; H
trast, the quantity, size and shape of the cable wires can be determined = height of the wires in the trough; and hi = depth of the ith layer.
in the preliminary design stage, and their huge axial forces will induce According to the geometric relationships [Fig. 5(a)-(b)], H and hi
considerable radial forces due to the bending transition, resulting in an can be obtained by
anchoring effect similar to gravity stiffness. The geometric configura-
tion of the cable wires in saddle thus is relatively certain and main- H = d + (n 1)· d ·cos (13)
tainable, presenting a discrete yet regular feature at a more microscopic
level.
hi = d/2 + (i 1)·d ·cos (14)
Therefore, it can be first qualitatively summarized that the existing
formula fails to properly take into account the restrictions of the
Janssen theory and the discrete nature of the cable wires, indicating 2. Top-side wire
that its theoretical basis is insufficient and its feasibility may have been L R
fs(1) = C2 f1,1 (15)
limited.

3. Analytical model and solution ev

f1,1 wv
+ f1,1
R
f1,1 =
Against the existing method, here emphasis is placed upon devel-
C1 + µ ws C2 (16)
oping an analytical approach of the cable-saddle system based on the
discrete contact relationships. First, the following three assumptions are R
made: fs(1) = C2 f1,Lm (17)

1. The cable wires in the saddle trough are regularly arranged as being
consistent with the conventional forming requirements [Fig. 5]. f1,evm + f1,wv
m
f1,Lm =
2. There is no lateral contact between adjacent wires in the same layer. C1 + µ ws C2 (18)
3. As with Assumption 3 in the Janssen theory, the friction between
where fs(Li ) and fs(Ri ) = lateral force on the left and right side surfaces,
contact surfaces is assumed to be fully mobilized and satisfies the
respectively, at the ith layer; µ ws = friction coefficient between wire
classical Coulomb’s law [19,27].
and saddle trough; and C2 = another refined transfer coefficient:
As illustrated in Fig. 5(a), a saddle containing three troughs divided C2 = sin µ w cos (19)
by two spacers or friction plates is considered. The cable in the center
trough is selected and reverted into discrete wires for analysis
3. Even-layer wire
[Fig. 5(b)]. The wire diameter is d, and the number of columns and
layers of the wires are m and n, respectively. Based on the boundary f iwv
,j
fiL, j = fiL 1, j + 1 +
conditions, the wires are categorized into eight types (see the legend). 2C1 (20)
Note that the filling of short wires is a common method in practice to
maintain the shape of the main cable in the saddle, Fig. 5 adds the case
of placing short wires in the side column of arbitrary layer. The contact fi,wv
j
forces between the discrete wires can be further clarified as Fig. 5(c) fiR, j = fiR 1, j +
2C1 (21)
depicted. Therefore, the equilibrium states of the eight representative
wires are acquired and presented in Fig. 5(d).
4. Side wire

3.1. Lateral force fs(Li ) = C2 (fiR,1 + fiL 1,1 ) (22)

Based on the lateral and vertical equilibrium conditions [Fig. 5(d)],

the internal forces of the cable-saddle system can be deduced as follow: fiL 1,1 (C1 wv
µ ws C2 ) + fi,1
fiR,1 =
C1 + µ ws C2 (23)
1. Top-inner wire

f1,evj + f1,wv
j
f1,Lj = f1,Rj = fs(Ri ) = C2 (fi,Lm + fiR 1, m 1 )
2C1 (10) (24)

C1 = cos + µ w sin (11)

4
L. Wang, et al. Engineering Structures 221 (2020) 111087

Fig. 5. Analytical model for the lateral force of cable-saddle system: (a) deployment of the saddle and cable; (b) discretization of cable wires; (c) force chains; and (d)
equilibrium conditions of representative wires.

fiR 1, m 1
(C1 µ ws C2) + fi,wv
m
fiL 1,1 (C1 µ ws C2 )
fi,Lm = fiR,1 =
C1 + µ ws C2 (25) C1 + µ ws C2 (28)

5. Odd-layer wire
fiR 1, m 1
(C1 µ ws C2 )
fi,wv fiL, m =
fiL, j = fi L
+
j C1 + µ ws C2 (29)
1, j
2C1 (26)
The formulations of lateral forces at this position are same as Eqs.
(22) and (24).
fwv 7. Bottom-inner wire
fiR, j = fiR 1, j +
1
2C1 (27) fnbv wv L
+ fnR
, j = f n, j + C1 (fn 1, j 1, j 1 ) (30)
where = bottom reaction force on the (n, j ) wire.
fnbv
6. Short wire ,j
8. Bottom-side wire

5
L. Wang, et al. Engineering Structures 221 (2020) 111087

fsL(n) = C2 fnL 1,1 (31)

fnbv wv L
,1 = f n,1 + C1 fn 1,1 (32)

fsR(n) = C2 fnR 1, m 1 (33)

fnbv wv R
, m = f n, m + C1 fn 1, m 1 (34)

From Eqs. (10)–(34), the lateral forces as well as other contact

forces can be acquired recursively. Accordingly, the total lateral force at
each side can be easily accumulated by
n
FyL = fs(Li)
i=1
n
FyR = fs(Ri )
i=1 (35)

3.2. Lateral pressure

As illustrated in Fig. 6, due to the adjacent side contact points are

close, the lateral pressure can be obtained through evenly distributing
the lateral force over the short height range

f y(L hi) = fsL(i ) / hi

f y(R hi) = fsR(i ) / hi
(36)
where and
fy(L hi) = lateral pressure on the left and right side sur-
fy(R hi)
faces, respectively, at depth of hi ; and hi = distribution height cor-
responding to the lateral force at the ith layer, as expressed by

d cos + d /2, i = 1 or n
hi =
2d cos , i = 2 ~ (n 1) (37) Fig. 7. Flow chart for the CSSDA model.

3.3. Model solution algorithm System Discrete Analytical (CSSDA) model in this study. The required
parameters can be properly determined in the preliminary design stage,
From the above theoretical derivations, an analytical model invol- and the CSSDA model can be efficiently analyzed by a top-down re-
ving multiple explicit formulas is proposed and named the Cable-Saddle cursive algorithm as shown in Fig. 7. In this paper, the software package
MATLAB is used [28]. Since the programming is straightforward, only
the main steps are described here. First, input the basic information
such as wire arrangement and saddle structure to generate the geo-
metric configuration of the cable-saddle system. Second, calculate the
contact forces of the wires in first layer from left to right (j increases
from 1 to m). If there is contact between wire and saddle side in this
layer, store the lateral force in the vector [ fy ]. Next, transfer the force
information of the first layer to the second layer, and calculate the
contact forces of the wires in this layer in sequence. By analogy, per-
form the calculation layer-by-layer from top to bottom (i increases from
1 to n), and finally output the concerned information such as the lateral
pressure and lateral force.
It should be noted that, for conciseness, the derivations are ex-
emplified by the case where the number of wire layers (n) is odd and
the top wires are fully arranged. Actually, when n is even, it is only
required to reverse the wire conditions of the odd layer and the even
layer in the above analysis; and when a part of the top layer is vacant, it
is only necessary to set the wire tension at this position to zero.
Moreover, since the key parameters (e.g., location and tension) are
embedded into each wire independently, the proposed CSSDA model is
more compatible and generalized.
Fig. 6. Schematic diagram for calculating lateral pressure.

6
L. Wang, et al. Engineering Structures 221 (2020) 111087

4. Error estimation assumed to be evenly shared by the bottom wires. The general ex-
pression of stacking angle is thus obtained as
Note that although the proposed analytical model is obtained from
(m 1) d + tot
strict mechanical relationships and logical thought experiments, the = arcsin
2(m 1) d (40)
effects of possible errors should be further clarified, providing a more
comprehensive evaluation of the CSSDA model. On the other hand, considering the most extreme case and the actual
limit, the maximum transverse gap can be determined by
4.1. Poisson effect
max
tot = pre + tols+ + m |tol w| + m
The cable wires will contract in radial direction under axial force, max
tot <d (41)
namely Poisson effect. Thus, a numerical study is performed involving a
cable wire with normal parameters: the wire diameter d = 5 mm, the where pre is the pre-widen space for ensuring a given number of wires
axial force Tw = 12 kN, the elastic modulus E = 210 GPa, and the can be placed into the trough, which is acquired by pre = m × tol w+ as
Poisson ratio = 0.3. Therefore, the radial strain can be calculated as previously mentioned; is the width deviation caused by the radial
contraction of single wire, i.e., = d × d , and d is calculated by Eq.
4Tw (38).
= 0.00087
d
E d2 (38) Eventually, by substituting Eq. (40) into Eqs. (11), (13), (14), (19)
The magnitude of less than one thousandth demonstrates that, for and (37), the influence of transverse gap can be taken into considera-
general engineering practice, the Poisson effect of the cable wires is tion by the CSSDA model.
weak. Nevertheless, this radial contraction will induce a tendency to
separate between adjacent wires in the same layer, which implies that 4.2.2. Influence analysis
the Assumption 2 used in the previous derivation is more likely to be A saddle trough containing multiple cable wires with a diameter of
real. As for the influence tendency of transverse gap caused by factors 5 mm is taken as an example to analyze the influence of transverse gap.
including such radial contractions, it will be analyzed in the following For completeness, the gap tot is set to a variable ranging from 0 to the
part. limit value 4 mm. The other parameters used in the analysis are listed in
Table 1. The friction coefficients are taken as 0.2, which is a more ac-
4.2. Transverse gap cepted value at present [1,16,14]. Based on the real cases (e.g.
[3,17,31]), the number of wire layers is appropriately expanded to 201
The saddle troughs and cable wires inevitably have positive or ne- to make the results more compatible. Similarly, the values of other
gative tolerances in manufacturing. To ensure that a given number of parameters involving the saddle trough and cable wires can also be
cable wires can be placed in the transverse direction of the saddle determined with reference to general engineering practice. By using the
trough on-site, the trough width should be designed to be sufficient in proposed algorithm, the core calculation process of this example takes
advance, e.g., by accumulating the positive tolerance of the wire dia- less than one second, showing high efficiency. The main configuration
meter [29,13]. Therefore, only positive width deviation (i.e., transverse of the used computer is CPU-Core i7-6700HQ-2.6 GHz and RAM-8 GB.
gap) may occur and needs to be studied. The variation of the lateral pressure is presented in Fig. 9, which
also includes the results of the existing formula. It is observed that, the
4.2.1. Analytical method lateral pressure calculated by the CSSDA model grows gradually as the
An analytical approach is illustrated in Fig. 8, with respect to a gap increases. This trend reflects the mechanism that the larger the
saddle trough containing two columns of cable wires. Suppose the total stacking angle caused by the gap, the more the direction of force chains
transverse gap is tot , then the stacking angle can be updated by Eq. (39) tends to be horizontal, resulting in the greater the lateral component of
according to the geometric relationships. the stress. For the existing formula, however, only slight opposite
changes are found at the upper and lower layers. This lower sensitivity
d + tot is because the gap mainly affects the stacking angle rather than the
= arcsin
2d (39) overall size (b and H), and unfortunately, the existing formula can only
Note that as the allowable tolerance of the width of saddle trough recognize the gap effects via b and H [recalling Eq. (9)].
(tols± = ± 0.15 mm) and the allowable tolerance of the diameter of cable The lateral pressure of the CSSDA model drops suddenly at the
wire (tol w± = ± 0.07 mm) are relatively small [30,29], tot can be bottom layer. This bottom effect, which has also been observed in a
simulated Janssen experiment [32,22], reveals the fact that the bottom
wires are supported by flat surface being different from the contact
cases of other wires. As proven in Section 2, the existing formula is
derived from the Janssen theory that has ignored the bottom effect, so
its curve hides this detail. In addition, a significant discrepancy between
the results of the existing formula and the CSSDA model can be found,
and the former is always larger.
To further quantify these issues, Table 2 compares the total lateral

Table 1
Parameters used in analysis of transverse gap.
Parameter Notation Value

Friction coefficients µ w , µ ws 0.2

Number of wire layers n 201
Number of wire columns m 10
Wire diameter (mm) d 5
Wire tension (kN) Tw 12
Radius of saddle trough (m) R 8
Fig. 8. Analytical model of transverse gap.

7
L. Wang, et al. Engineering Structures 221 (2020) 111087

Fig. 10. Influence of transverse gap on total lateral force.

5.1. Wire diameter

After the cable tension is determined at the initial design stage, the
diameter of cable wires should be selected, normally within a range of
Fig. 9. Influence of transverse gap on lateral pressure. 4.5–5.5 mm [11,29]. Different selections of the wire diameter may af-
fect the lateral force, a numerical study is thus performed. The case of
forces. Obviously, the bottom effect is slight (around 0.86%) and can be using the parameters in Table 1 is first taken as the benchmark, and two
disregarded in actual engineering. The existing formula seriously additional cases of d = 4.5 mm and d = 5.5 mm are obtained according
overestimates the lateral force, e.g., the maximum deviation reaches to the equivalence principle that the total sectional area and total
35.47%, indicating the unsuitability of the existing formula. As the gap tension of cable wires keep unchanged. Accordingly, Fig. 11 depicts the
expands, the lateral force of the CSSDA model increases, while that of results of lateral pressure and total lateral force.
the existing formula decreases, but both develop linearly (Fig. 10). It is obvious that the wire diameter has influence on the lateral
When tol w± and tols+ adopt their allowable values, tot
max
= 1.59 mm can be pressure, especially for the lower layers. The smaller the wire diameter,
obtained by Eq. (41). Further, the dashed line in Fig. 10 displays that the smaller the lateral pressure at the same depth. However, as the
the corresponding impact rate is 4.08%, demonstrating that the gap has inserted table shown, the total lateral force (Fy ) increases with the de-
a limited enlargement effect on the lateral force. crease of wire diameter. This opposite effect reveals that the magnitude
Therefore, the proposed CSSDA model provides a possibility to es- of lateral force mainly depends on the dimension of height, i.e., a
timate the gap effect, yet in real practice, engineers can make reason- smaller diameter induces a larger height, resulting in a greater lateral
able trade-offs of the effect according to the application object. For force. Based on this understanding, the wire diameter can be selected
example, when checking the structural strength of the saddle, it is purposefully in practice, e.g., the relatively thin wire is preferred when
suggested to faithfully reflect the gap effect; whereas when evaluating the cable-saddle system requires a great anti-slip capability. Moreover,
the anti-slip capacity of the main cable, the small beneficial effect of the it is remarkable that the change rate of the total lateral force is close to
gap can be conservatively neglected. that of the wire diameter by 10% in both cases. Therefore, the influence
of wire diameter can be roughly quantified in this way.

5. Sensitivity study 5.2. Wire arrangement and friction coefficient

Based on the proposed model, this section will conduct sensitivity The practical interest of this part is as follows. First, after de-
studies to clarify the influence of some key factors, including diameter termining the specification of cable wires, it is necessary to design the
and arrangement of the wires, friction coefficient and possible extra wire arrangement, which is mainly reflected in the aspect ratio. Second,
pressure. although it is currently accepted that the friction coefficient of 0.2 is
closer to the actual situation, the recommended value of 0.15 is still

Table 2
Comparison of total lateral force.
Gap (mm) Calculation method Total lateral force Fy (kN/m) Impact rate on Fy (%)
Full height Without bottom Bottom Calculation method Gap

0 Existing formula 4381.1 — — 35.47 —

CSSDA 3233.9 3205.6 0.88 — —
1 Existing formula 4324.1 — — 30.30 −1.30
CSSDA 3318.5 3289.7 0.87 — 2.62
2 Existing formula 4267.1 — — 25.44 −2.60
CSSDA 3401.6 3372.2 0.86 — 5.19
3 Existing formula 4210.0 — — 20.86 −3.91
CSSDA 3483.3 3453.4 0.86 — 7.71
4 Existing formula 4152.7 — — 16.53 −5.21
CSSDA 3563.6 3533.1 0.86 — 10.20

8
L. Wang, et al. Engineering Structures 221 (2020) 111087

Fig. 13. Changes of total lateral force with wire arrangement and friction
coefficient.

m > 50 (approximates to b/ H > 0.3), the results only have a weak

Fig. 11. Influence of wire diameter on lateral pressure and total lateral force. growth. Therefore, the range of wire arrangement that can produce a
significant effect on the lateral force is relatively small.
The results of the existing formula are always larger, but the dis-
conservatively adopted in general engineering practice, indicating that
crepancy is not obvious when m is small. However, a slight increase of
the friction coefficient currently used is between 0.15 and 0.2.
m will trigger a rapid expansion of that discrepancy. For example, when
Therefore, in a specific application, there is a choice space of the wire
m merely increases to 10 (approximates to b/ H = 0.06), the distribution
arrangement and the friction coefficient, and different choices would
of lateral pressure between the two has shown a significant difference,
affect the lateral force.
and the deviation of total lateral force has reached 30%. This dis-
The parameters used in this part are consistent with Table 1, except
crepancy is mainly because the existing formula inherits the limitations
that the number of wire columns (m) and the friction coefficients
of the Janssen theory that applies only to deep containers.
( µ w , µ ws ) are set as variables. Note that since these two friction coef-
On the other hand, it is found that the lateral force has a negative
ficients have not been distinguished in real practice thus far, the two are
correlation with the friction coefficient, and the saturation of lateral
taken as the same herein. The results of lateral pressure and total lateral
force is more abrupt as the friction coefficient is higher. Nevertheless,
force are illustrated in Fig. 12 and Fig. 13, respectively.
when the friction coefficient varies within the normal range of 0.15 to
The distribution features of the lateral pressure obtained by the two
0.2 [8,7,17], the distribution and variation characteristics of the lateral
approaches are basically consistent, i.e., as m increases, the distribution
force are scarcely affected.
pattern gradually changes from negative exponential type to near-linear
type. Both the lateral pressure and the total lateral force increase with
the increase of m, but their increments decrease gradually, e.g., when

Fig. 12. Changes of lateral pressure with the number of wire columns (m), when the friction coefficients are equal to: (a) 0.2; and (b) 0.15.

9
L. Wang, et al. Engineering Structures 221 (2020) 111087

Fig. 14. Influence of extra pressure on the distribution of lateral pressure, in cases of different wire arrangements: (a)-(c) m = 5, 10, 20, respectively, when n = 201;
and (d)-(f) m = 5, 10, 20, respectively, when n = 200 .

5.3. Extra pressure the transmission loss of extra pressure due to the packing friction,
which is similar to the mechanical behavior found in bulk stocks.
To enhance the stability between the cable and saddle, extra pres- In contrast, when n is even [Fig. 14(d)-(f)], the lateral pressure is in
sure can be applied on the top of cable within the saddle, e.g., by a ‘zig-zag’ distribution, and the step range always includes (m 1) side
pressing a cover with prestressed bolts against the cable [11,17]. The contacts. Therefore, the significant yet regular effects of the parity of
proposed CSSDA model is therefore used in this part to clarify the in- the number of wire layers on the lateral pressure are revealed.
fluence of the extra pressure on the lateral force. Meanwhile, an ad- On the other hand, Fig. 15(a) depicts the changes of total lateral
jacent case of n = 200 is adopted to study the effects of the parity of the force with extra pressure. An obvious linear relationship can be found
number of wire layers. First, the results of lateral pressure are illu- that is independent of the wire arrangement. However, a wider ar-
strated in Fig. 14, where ‘Ref’ denotes a reference case without any rangement induces a greater effect of the extra pressure as can be seen
extra pressure; ‘+ x%’ denotes a case that the applied extra pressure from the curve slopes. Additionally, it is interesting to note that al-
equals x% of the vertical pressure (Fv ) generated by the cable itself; and though the parity of the number of wire layers has obviously affected
‘Equ’ represents a case that Fv is equivalently converted into the extra the lateral pressure, the total lateral force is hardly affected. This
pressure. phenomenon can also be observed in Fig. 15(b), where the extra pres-
The results show that the extra pressure has significant effects on sure is equivalently converted from the self-pressure. Meanwhile, the
the magnitude and distribution of the lateral pressure. When the logarithmic change indicates that the equivalent pressure will not fur-
number of wire layers (n) is odd [Fig. 14(a)-(c)], the extra pressure ther enhance the lateral force when the wire arrangement is widened to
makes the lateral pressure first jump sharply and then present a pulse- a certain extent.
like change. The pulse frequency decreases as m increases, yet the step
range always includes (m 2) side contacts. When the extra pressure is
6. Case study
large and m is small, the lateral pressure tends to decrease along the
depth; otherwise, the overall tendency of lateral pressure is close to the
6.1. Description of actual structure
reference case. However, in the cases of ‘Equ’, the lateral pressure ex-
hibits a nonlinear attenuation along the depth, especially when the
The Taizhou Yangtze River Bridge [1,31] in China is selected to
wires are narrowly arranged [Fig. 14(a)]. This feature actually reflects
exemplify the application of the proposed CSSDA model. As depicted in

10
L. Wang, et al. Engineering Structures 221 (2020) 111087

Fig. 15. Changes of total lateral force: (a) with different amounts of extra pressure; and (b) with the number of wire columns when applying extra pressure
equivalently.

Fig. 16, this bridge is a three-tower suspension bridge with an inverted

Y-shaped steel middle tower. The main cable is composed of 169
bundles of strands, and each strand consists of 91 parallel galvanized
steel wires that are 5.2 mm in diameter. The strands are separated into
15 columns in the middle saddle (Fig. 17), and the radius of the saddle
trough is 5.92 m. To increase the frictional resistance, two vertical
friction (VF) plates with a thickness of 60 mm are installed into the
middle saddle.
Note that as the thick VF plates are firmly fixed on the saddle, the
saddle can be regarded as a parallel system that is composed of three
sub-saddles, i.e., one middle sub-saddle and two side sub-saddles
(Fig. 17). Each sub-saddle contains 5 columns of strands divided by 4
spacers. Because of the unreliable durability of the spacers, their anti-
slip effects are ignored being consistent with real practice [9,11,7,17].
The friction coefficient of contact interfaces used in calculation is a
traditional value of 0.15 [13]. From the conventional finite element Fig. 17. Deployment of the middle saddle and strand wires.
analysis, the tension of each main cable in the initial dead load state is
193,153 kN, which is adopted to calculated the cable-saddle frictional The total frictional resistance of the main cable comprises three
resistance without considering the small effect of the live load [7,3,17]. components: the bottom friction accounts for the principal part
(67.8%), the saddle sides provide the least part (7.4%), and the two VF
6.2. Analysis of frictional resistance plates contribute 24.8%. For general engineering, the contact surface
between the cable and the bottom or side of the saddle is substantially
According to the lateral equilibrium, the trough with the most defined, implying that it is hard to further enhance the frictional re-
strands in a sub-saddle can be selected as a reference for calculating the sistance on these surfaces. In contrast, each VF plate can provide two
lateral pressure of this sub-saddle (Fig. 17). Then, the lateral force in- additional lateral surfaces for the cable, and in principle, more VF
teracting with the VF plate or the saddle can be obtained by integrating plates can be deployed in some or even all spaces between the strand
the known lateral pressure along the corresponding contact height. As columns. Therefore, it can be inferred that the lateral friction is pro-
for the radial force acting on the bottom of each sub-saddle, it can be mising to improve the anti-slip capacity of the cable-saddle system, and
easily calculated by referring to Eq. (4). After all the normal forces are the use of VF plates is an effective way to fully utilize the anti-slip
obtained, the corresponding frictional resistance can be calculated ac- potential of the lateral friction.
cording to Coulomb’s law. The computation time of using the proposed The frictional resistance contributed by the VF plates in the middle
model is about 1.7 s, excluding the time before and after processing. sub-saddle is 1012.4 kN/m, while that in the side sub-saddles is 778.3
The results are listed and compared in Table 3, where the frictional kN/m. That is, although the VF plates provide the same number of
resistance of side sub-saddle actually represents that of the two side surfaces for the middle and side sub-saddles, the frictional resistance
sub-saddles, and denotes the frictional resistance of a surface as a contributed by the VF plates is different between the sub-saddles. This
percentage of the total value. difference is jointly caused by the differences of the reference trough

Fig. 16. Elevation view of the Taizhou Yangtze River Bridge (unit: m).

11
L. Wang, et al. Engineering Structures 221 (2020) 111087

Table 3 suspension bridges. Besides, the comparison again confirms that the
Frictional resistance and its components. existing formula is insufficient for evaluating the anti-slip safety of the
Calculation Item Sub-saddle Bottom Side surface Total main cable. Through the above comprehensive analysis, the sig-
method surface nificance of accurately calculating the lateral forces and adequately
VF plate Saddle utilizing the lateral friction is reflected more intuitively.
CSSDA Frictional Middle 1998.2 1012.4 0.0 3010.6
resistance Side 2895.9 778.3 532.7 4206.9
7. Conclusions
(kN/m) Total 4894.1 1790.8 532.7 7217.5
Percentage 67.8 24.8 7.4 — On the basis of the above investigations, the main findings can be
(%) summarized:
Frictional Middle 1998.2 1294.4 0.0 3292.6
Existing resistance Side 2895.9 1014.1 708.2 4618.2
formula (kN/m) Total 4894.1 2308.5 708.2 7910.8
1. The existing formula is a simple evolution from the Janssen theory
Percentage 61.9 29.2 9.0 — and fails to faithfully reflect the physical and mechanical char-
(%) acteristics of the cable-saddle system. The lateral forces calculated
by the existing formula are always larger, indicating its insufficient
a
Discrepancy Middle 0.0 27.8 0.0 9.4
(%)
applicability, especially for evaluating the frictional resistance of the
Side 0.0 30.3 33.0 9.8
Total 0.0 28.9 33.0 9.6
main cable.
2. The presented analytical model, taking into account the discrete
a
Discrepancy between the results obtained by the two calculation methods. contact nature of the main cable within the saddle, offers an op-
portunity to the calculation of lateral force and the utilization of
Table 4 lateral friction, which is of great significance for the design and
Evaluation of the anti-slip safety of main cable. promotion of multi-tower suspension bridges.
Calculation method Ff (kN/m) Fb (kN/m) µn K
3. The increase effect on lateral force that the transverse gap of saddle
trough may have is limited. Using a smaller diameter of cable wires
CSSDA 7217.5 32627.1 0.221 2.63 leads to a greater lateral force. With the number of wire columns
Existing formula 7910.8 32627.1 0.242 2.89 increasing, the lateral force will rapidly grow to a stable value, and
Without lateral friction 4894.1 32627.1 0.150 1.79
the distribution of lateral pressure also changes from a negative
exponential form to a near-linear form. The lateral force is nega-
tively correlated with the friction coefficient, whereas varying the
and the contact range, i.e., the more strands in the reference trough and
friction coefficient within a normal range (0.15–0.2) hardly affect
the larger the contact range, the greater the lateral frictional resistance.
the overall characteristics of the lateral force. The extra pressure has
Thus, the total anti-slip effects of VF plate cannot be accumulated
an obvious effect on the distribution of lateral pressure, and this
simply according to the number of VF plates.
effect is closely associated with the parity of the number of wire
In addition, there are obvious discrepancies between the results of
layers. Nevertheless, the lateral force always has a linear relation-
the two calculation methods, e.g., the discrepancy of the total frictional
ship with the extra pressure.
resistance is 9.6% and that of the lateral frictional resistance is nearly
4. The case example shows that, the lateral friction could increase the
30%. It is therefore unsafe to use the existing formula to calculate the
cable-saddle frictional resistance by 32.2%, incorporating with two
anti-slip friction of the cable-saddle system, especially in case of de-
VF plates that contribute 24.8%. It is thus intuitively confirmed that
ploying more VF plates.
the lateral friction has considerable application potential for en-
hancing the anti-slip capacity of cable-saddle system, and the use of
6.3. Evaluation of anti-slip safety VF plates is an effective approach to adequately exploit this poten-
tial.
The nominal friction coefficient ( µ n ) is usually used as an indicator
of the frictional resistance in practice Declaration of Competing Interest
F
µn = f
Fb (42) None.

where Ff = total frictional resistance; and Fb = radial force acting on

Acknowledgments
the bottom of saddle.
Thus, the anti-slip safety factor K can be expressed as
This authors would like to appreciate the support of the Science and
µn Technology Program of Zhejiang Provincial Communication
K=
µcr (43) Department [Grant No. 2011H03], the Project Funded by China
Postdoctoral Science Foundation [Grant No. 2020M672634], the State
where µcr = friction coefficient corresponding to the critical slip state, Key Laboratory of Traction Power at Southwest Jiaotong University
which can be calculated from the Euler formula [9,7,17]. For this [Grant No. TPL1802], and the National Natural Science Foundation of
bridge, µcr = 0.084 . China [Grant Nos. 51678247, and 51878295].
Based on Eqs. 42,43 and the known data in Table 3, the evaluation
results are acquired and listed in Table 4. When the lateral friction is References
properly taken into account, the anti-slip safety factor K reaches 2.63,
indicating that the main cable has sufficient frictional resistance. [1] Ge Y, Xiang H. Extension of bridging capacity of cable-supported bridges using
However, when following the traditional simplified method that ig- double main spans or twin parallel decks solutions. Struct Infrastruct E
nores the lateral friction, K = 1.79 fails to meet the requirements of 2011;7(7–8):511–67.
[2] Thai HT, Choi DH. Advanced analysis of multi-span suspension bridges. J Constr
K 2 in the specification [13]. It is thus proved that, the cable-saddle Steel Res 2013;90:29–41.
frictional resistance has been seriously underestimated by the tradi- [3] Cheng Z, Zhang Q, Bao Y, Jia D, Bu Y, Li Q. Analytical models of frictional resistance
tional method, which is exactly the reason for the realistic dilemma in between cable and saddle equipped with friction plates for multispan suspension
bridges. J Bridge Eng, vol. 23, 1.
the innovative design and reasonable promotion of the multi-tower

12
L. Wang, et al. Engineering Structures 221 (2020) 111087

[4] Zhang W, Yang C, Wang Z, Liu Z. An analytical algorithm for reasonable central plates. J Bridge Eng, vol. 24, 4.
tower stiffness in the three-tower suspension bridge with unequal-length main [18] Janssen HA. Versuche über getreidedruck in silozellen. Z Ver Deutch Ing
spans. Eng Struct 2019;199. 109595. 1985;39(35):1045–9.
[5] Nazir C. Multi-span balanced suspension bridge. J Struct Eng [19] Schillinger D, Malla RB. Analytical elastic solution based on fourier series for a
1986;112(11):2512–27. laterally confined granular column. J Eng Mech 2008;134(11):937–51.
[6] Forsberg T. Multi-span suspension bridges. Int J Steel Struct 2001;1(1):63–73. [20] Goodey RJ, Brown CJ, Rotter JM. Rectangular steel silos: Finite element predictions
[7] Zhang Q, Cheng Z, Cui C, Bao Y, He J, Li Q. Analytical model for frictional resistance of filling wall pressures. Eng Struct 2017;132:61–9.
between cable and saddle of suspension bridges equipped with vertical friction [21] Ooi JY, Rotter JM. Wall pressures in squat steel silos from finite element analysis.
plates. J Bridge Eng, vol. 22 (1). Comp Struct 1990;37(4):361–74.
[8] Takena K, Sasaki M, Hata K, Hasegawa K. Slip behavior of cable against saddle in [22] Ovarlez G, Clément E. Elastic medium confined in a cylinder versus the janssen
suspension bridges. J Struct Eng 1992;118(2):377–91. experiment. Eur Phys J E 2005;16:421–38.
[9] Hasegawa K, Kojima H, Sasaki M, Takena K. Frictional resistance between cable and [23] Malla RB, Schillinger D, Vila LJ. Experimental determination of friction coefficient
saddle equipped with friction plate. J Struct Eng 1995;121(1):1–14. and mobilization force for a laterally confined granular column. Granul Matter
[10] Eurocode-3, Design of steel structures–Part 1-11: Design of structures with tension 2014;16(6):843–55.
components, European Committee for Standardization, Brussels (Belgium); 2006. [24] Nedderman RM. Statics and kinematics of granular materials. Cambridge, U.K.:
[11] Gimsing NJ, Georgakis CT. Cable supported bridges: Concept and design. 3rd ed. Cambridge University Press; 1992.
New York: John Wiley & Sons; 2013. [25] Feng Y, Hua Y. Modified janssen theory for flexible circular bins. J Struct Eng
[12] Coulomb CA. Sur une application des régles de maximis & minimis à quelques 1996;122(4):454–6.
problèmes de statique relatifs à l’architecture. Mém Math Acad R Sci [26] Chen JF, Rotter JM, Ooi JY. A review of numerical prediction methods for silo wall
1776;7:343–82. pressures. Adv Struct Eng 1998;2(2):119–36.
[13] JTG/T D65-05-2015, Specifications for design of highway suspension bridge, [27] Montoya A, Waisman H, Betti R. A simplified contact-friction methodology for
Ministry of Communications and Transportation of People’s Republic of China, modeling wire breaks in parallel wire strands. Comp Struct 2012;100–101:39–53.
Beijing; 2015. [28] MATLAB user’s guide, MathWorks Inc., Natick, Mass.; 2015.
[14] Zhang Q, Kang J, Bao Y, Cheng Z, Jia D, Bu Y. Numerical study on cable-saddle [29] Meng F. Suspension bridge. Beijing: China Communications Press; 2011. (in
frictional resistance of multispan suspension bridges. J Constr Steel Res Chinese).
2018;150:51–9. [30] GB/T 17101-2008, Hot-dip galvanized steel wires for bridge cables, Standardization
[15] Zhang Q, Guo H, Bao Y, Cheng Z, Jia D. Antislip safety of double-cable multispan Administration of People’s Republic of China, Beijing; 2008.
suspension bridges with innovative saddles. J Bridge Eng 2020;25(5):04020021. [31] Jiangsu Provincial Highway Planning and Design Institute (JPHPDI), China Major
[16] Wang L, Shen R, Bai L, Wang Y. Test for slip behavior and mechanical character- Bridge Reconnaissance and Design Institute (CMBRDI), Architecture Design and
istics between main cable and saddle in suspension bridges. China J Highway Research Institute of Tongji University (ADRITJU), Documents of construction de-
Transp 2018;31(9). pp. 75–83+103, (in Chinese). sign of Taizhou Yangtze River Bridge; 2007 [in Chinese].
[17] Wang L, Shen R, Wang C, Zhang S, Wang Y. Theoretical and experimental studies of [32] Ovarlez G, Fond C, Clément E. Overshoot effect in the janssen granular column: A
the antislip capacity between cable and saddle equipped with horizontal friction crucial test for granular mechanics. Phys Rev E 2003;67: 060302/1 – 060302/4.

13