International Journal of Mechanical Sciences 113 (2016) 196–210

Contents lists available at ScienceDirect

International Journal of Mechanical Sciences

journal homepage: www.elsevier.com/locate/ijmecsci

A simplified finite element model for structural cable bending

mechanism
Yujie Yu b, Xiaoxiang Wang b, Zhihua Chen a,n
a
State Key Laboratory of Hydraulic Engineering Simulation and Safety, Tianjin University, China
b
Department of Civil Engineering, Tianjin University, Tianjin 300072, China

art ic l e i nf o a b s t r a c t

Article history: A simplified finite element modeling method for structural cables on bending and wire sliding problems
Received 5 October 2015 are proposed and formulated. The model is characterized by a beam-spring composition that uses helically
Received in revised form arranged short-beam elements to simulate spiral wires, radially-placed rigid beams to replace wire sections
4 March 2016
and spring elements to capture contact extrusion and friction between adjacent wires. Winding and
Accepted 6 May 2016
contacting patterns for semi-parallel wire cables and spiral strands are discussed and then used to for-
Available online 11 May 2016
mulate the model establishments. In semi-parallel wire cable, any selected wire section can be simplified
Keywords: as a seven-node set to represent the center wire and the surrounded six contact spots. While in spiral
Finite element modeling method strand, wire section can be replaced by four rigid beams that connect the four contact points to wire center.
Cable bending
Helical wire compression, sheath pinch effect and contact friction are all considered and calculated, and
Semi parallel wire cable
then bending simulations are conducted on large dimension cables with different end conditions, wire
Galfan spiral strand
Contact friction constructions and pretension levels. Simulation results are compared to the test data, and prove that the
proposed model can well predict the bending behavior and the varying trend of flexural stiffness. Both
cables all express a bilinear elastic–plastic like flexural strength–deflection response. Semi-parallel wire
cable has sinusoidal deflection out of the bending plane, and the range of bending gradient is influenced by
the cable end conditions. Galfan strand shows little unwinding and torsion effect due to the canceling out
effect from alternative twisting between adjacent layers. The model can efficiently simulate long and large
dimension cables in a short time, implying its promising applicability in cable studies.
& 2016 Elsevier Ltd. All rights reserved.

1. Introduction applications often impose some transverse loads or curvature like

wind, water flow or hangers. Such perturbations may cause stress
Cables are composed of assemblages of wires that are helically concentrations in cable wires, especially near the ends or termina-
wrapped around a core, which provide high longitudinal strength and tions, with a consequent danger of fatigue failure [3].
minimal lateral stiffness. Cables often function as structural elements Due to the complex construction, deforming mechanisms of
in different structure fields, as semi-parallel wire cables in bridges, cable, especially the bending behavior, are difficult to understand.
strands in cable-supported structures, wire ropes in conveyance, and Different approaches have been explored to identify and predict the
conductor cables in electrical substations and transmission lines. Due characteristics of bent cable. Several previously developed bending
to the helical wires wrapped construction, the mechanical behavior of simulations [4-7] and dynamic tests [8,9] reveal that actual strands
cables under tensioning and bending is complex, especially given the showed an intermediate performance between parallel-arranged
potential frictional sliding between wires in cables. The effective wire bundle and solid beam mode as a result of wire slips in the
bending stiffness of cable may vary between two extreme states that cable. However, test data are scattered due to the inherently com-
correspond to fully stuck and fully slipped wires, and these states can plex mechanism, high testing and measurement requirements, and
differ by as much as two orders of magnitude [1]. Furthermore, the different choices of testing methods. Given the difficulties and un-
construction of cable would exhibit an extremely complex inner certainties in cable testing and measurements, theoretical deduc-
stress state within the assemblage of wires [2]. Although strands and tions were also proposed to describe this bending behavior.
cables are primarily used for axial loading capacity, practical Nevertheless, consistent conclusions are still difficult to reach.
Contracted results may even be generated as a result of dissimilar
n
Corresponding author. basic assumptions and simplifications. In Lanteigne's models [10],
E-mail address: zhchen@tju.edu.cn (Z. Chen). slipping begins from the outermost wires, whereas wire slip starts

http://dx.doi.org/10.1016/j.ijmecsci.2016.05.004
0020-7403/& 2016 Elsevier Ltd. All rights reserved.
 Y. Yu et al. / International Journal of Mechanical Sciences 113 (2016) 196–210 197

near the neutral axis in Papailiou's model [6] and in Raoof's test spring elements. In this study, several basic assumptions are
observations [4]. Nonetheless, these studies mainly focus on elec- adopted:
trical conductors, with not much discussion on structural cables.
Finite element (FE) modeling has offered a means of predicting the 1. Wire sections are supposed to follow the plane section princi-
friction condition and behavior of bent cables in recent years [11,12]. ple; thus, the shear deformation in each wire is ignored.
However, numerical simulations are often employed with full-scale 2. Sliding is supposed to occur only along the longitudinal direc-
modeling and fine mesh with 8-node brick elements and contact tion. Coulomb friction is adopted for the contact friction calcu-
elements; thus, numerous elements are generated and the calculating lation with the friction coefficient μ.
magnitude is considerable [13]. To diminish the calculation amount 3. Plasticity development for steel wires is not considered in this
and to ensure convergence, such FE models are often limited within a study. All wires are made of the same material and remain
short piece [14] or are simply built with no wire slippage assumptions elastic during bending.
[15,16]. As a result, overall behavior is difficult to determine, and then
the accuracy and detailed bending mechanisms of long cables are hard Under these assumptions, the cable can be characterized by a
to get. Other theoretical studies involve complex equations and de- beam spring composition. Wires are simplified into series of he-
ductions that are challenging to understand and hard to be applied for lically arranged, short-beam elements that are positioned along
in-depth studies [1,17,18]. Chen [19,20] recently proposed a simplified helical wire axes to achieve a slight twisting angle (represented by
finite beam element method to analyze the thermal expansion me- the red, pink, and green lines in Fig. 1(a). These elements are
chanism of cables. This method simulates the twisting of wires using hereafter referred to as “wire beams”). Six short beams with a
helically arranged short beams, which effectively solves the thermal length as the wire radius, and a large modulus are placed diver-
expansion problem experienced with steel cables. gently along the wire section (denoted by the blue lines in Fig. 1(a).
In the current study, an advanced FE modeling method is These beams are hereafter called “rigid beams”) to ensure the
proposed for the bending and wire sliding problems of structural plane section assumption for helical wires. Nonlinear spring ele-
cables. The twisted wires, contact conditions, and stick–slip state ments are positioned at the outer end of the divergently dis-
in the cable are considered in this model. Wire contact compres- tributed rigid beams. These spring elements represent contact
sion and the corresponding friction resistance are also calculated spots (the red spots in Fig. 1(a)) and are used to integrate contact
and incorporated into the simulation. Moreover, wire contacting extrusion and friction resistance. This method converts line con-
pattern of semi-parallel wire cables and spiral strands are dis- tact into point interactions at the ends of the rigid beams. Three
cussed and are used to formulate the model establishment pro- spring elements are placed at the contact spots: two springs (the
cesses. Cables with different dimensions, wire constructions, end X,Y springs) along the wire section plane to simulate the extrusion
conditions, presence, and pretension magnitudes are all simulated, effect, and another spring (the Z spring) along the longitudinal
and then the results are compared and validated with experi- direction of cable to govern the contact friction. Therefore, friction
mental data. Finally, the flexural stiffness, deflection patterns and resistance can be calculated based on the sum force of X,Y springs,
wire force distributions are carefully studied and compared for whereas real-time friction and slip can be obtained according to
both semi-parallel wire cables and spiral strands. the force and elongation of the Z spring. An FE model of a seven-
wire strand is presented in Fig. 1(b), in which element dimensions
are displayed as helically arranged wire beams. The black lines and
red points represent the rigid beams and spring elements,
2. Formulation of the advanced nonlinear finite element
respectively.
model

2.1. Basic concept of the advanced finite modeling method 2.2. Selection of elements in the cable model

The proposed FE model simulates twisted wires and The advanced nonlinear FE models are established and com-
inter-wire interactions through specially arranged beams and puted using the commercial finite element software ANSYS.

(a) Arrangement of elements in the FE modelling method; (b) 3D display of a seven-wire strand model
Fig. 1. Formulation of elements and example of a FE model. (For interpretation of the references to color in this figure, the reader is referred to the web version of this
article).
198 Y. Yu et al. / International Journal of Mechanical Sciences 113 (2016) 196–210

Element PIPE16 is adopted for the wire beam elements (elastic

modular 2.1
1011). PIPE16 is a uniaxial element with tension–
compression, torsion, and bending capabilities. This element is
based on 3D beam element theory and comprises six freedoms at
each node. The rigid beam is built with PIPE16 as well and pos-
sesses the same cross-section geometry as the wire beam, but has
a higher elastic modular value of 2.1
1016 to ensure rigidity.
COMBINE39 is a unidirectional element with nonlinear gen-
eralized force–deflection capability and used for the spring ele-
ments. This element is often employed in simulating the sliding
conditions between steel bars and concrete of reinforced concrete
structures. Nonlinear force–elongation (F–E) relations must be
defined in advance, as shown in Fig. 2. In this study, the X, Y
springs are assumed to have a proportional linear relation with the
stiffness as the steel modular. Meanwhile, the Z spring has a multi-
linear relation with the ultimate strengths at each load step cal-
Fig. 2. F–E relationship of the spring element (COMBINE39) along the X, Y, and Z
directions.
culated on the basis of the X, Y spring forces.

(a) Formulation of the FE model and contact conditions for the semi-parallel wire cable

(b) Formulation of the FE model and contact conditions for the spiral strand
Fig. 3. Finite element modeling and contact conditions for semi-parallel wire cable and spiral strand. (For interpretation of the references to color in this figure, the reader is
referred to the web version of this article).
 Y. Yu et al. / International Journal of Mechanical Sciences 113 (2016) 196–210 199

2.3. Establishment of the finite element cable model 3. Pre-calculations and the friction resistance determination
method
The construction process of cable model is illustrated in Fig. 3.
First, seven nodes are assigned: one is located in the center while 3.1. Pinch effect of polyethylene protection
the other six are distributed evenly along the wire boundary to
represent the entire wire section. Rigid beam elements connect Hot extrusion polyethylene (PE) protection is generally performed
the center node to the other six nodes to generate a seven-node outside a parallel or a semi-parallel cable to hold wires together and
set that satisfies the plane section assumption and then represents to protect the cable from erosion. Then, a PE sheath induces a pinch
the entire wire section. The segment is then circularly copied effect that applies circular compression on the cable and partly
contributes to the contact extrusion among internal wires (Fig. 4(a)).
around the center wire to build the surrounding six wire sections.
This compaction originates from the cold contraction effect of
If directed properly, any two adjacent beam-node sets would each
the PE sheath during fabrication. Therefore, we can obtain the
possess one node located at contacting region. Three spring ele-
following equation based on the thermodynamic principle:
ments along different directions can then be built between the
two coincident nodes to simulate the three direction contact in- 1 Δv
β=
teractions. In Fig. 3, three spring elements are placed at each red v Δt (1)
point: the X and Y springs obtain contact extrusion, and the Z The strain produced as a result of thermal volume change is
spring represents the sliding states. Fig. 3(a) describes the con- denoted by β × Δt × h, and the corresponding thermal stress Tt is
struction pattern of semi-parallel wire cable that has all the wires determined as follows:
wrapped toward the same direction. Besides, the contact spots
Tt = EβΔt h (2)
shift along helical wires with their relative positions keeping the
same. When the seven-node sets and spring elements for all the where E is the elastic modulus; β is the thermal expansion coef-
wire sections are established, duplicate all of these components ficient; Δt is the change in temperature; Δv is the volume incre-
along cable axes (Z-axis) with a small rotation about Z-axis to ment; Tt is the corresponding thermal stress; h is the thickness of
obtain other cable sections. Then connect the wire section centers the sheath; and q0 is the interface contraction of the PE sheath to
successively with PIPE16 (wire beam element) to produce helically the wires. We then establish:
positioned wire beams. π π

Spiral strand is another generally used cable that has different

Tt = EβΔth = ∫0 sin θ⋅q0 dl/2 = ∫0 sin θ⋅q0 r0 dθ /2 = q0 r0.
(3)
wire construction pattern. Such strands display alternating windings,
Assuming that pinch force is divided by the central angle oc-
thus inducing a composite of line contact (between wires in the same
cupied by each wire at the outermost layer, contraction stress is
layer) and point contact (between wires at adjacent layers) modes. In
computed as follows:
the establishment of a spiral strand model, the wire sections in ad-
jacent layers are duplicated in opposite directions. In Fig. 3(b), nine q0 li = q0 r0 θi = EβΔt h θi (4)
wires in two adjacent layers are selected and the contact spots in
Where θi is the radius angle shared by selected wire i at the
between are marked in the left-hand plot. The contact spots alternate outermost layer; In the FE model, sheath compression can be
in terms of numbers and locations for selected wires between ad- simplified as the centripetal force applied to the wire beam nodes
jacent cable-sections. This pattern can also be reflected in the plot of of the outermost layer. Moreover, the value of this simplified load
the entire cable piece (middle) and in the 3D FE model (right). When is the sum of the pinch effects in the surrounding area, as in-
two wires are point contacted at one cable section, then at next cable dicated in Fig. 4(b).
section the contact locations and even the two contact wires will get
changed. Therefore, the key step in building a spiral strand is to locate 3.2. Clench effect of tensioned spiral wires
the contact spot locations and built coincident node pair and spring
elements. The detailed contact finding and model establishing pro- The wire in the spiral strand and in the semi-parallel wire cable
cess for spiral strand will be illustrated in detail below. exhibits a helix geometry, which facilitates clenching compaction

(a) Pinch effect of PE protection (b) Simplification of the sheath pinch effect
Fig. 4. Sheath pinch effect consideration.
200 Y. Yu et al. / International Journal of Mechanical Sciences 113 (2016) 196–210

through the following equation:

⇀ 2π⋅T
q ⋅Δs = ( −r cos t + r cos ( t + Δt ), r sin ( t + Δt ) − r sin t , 0)
(2πr ) 2 + L2

2 2 πTr
= 1 − cos (Δt ) .
(2πr ) 2 + L2 (9)

Then take the limit as Δs goes to infinity:

→ 2 2 π⋅Tr 1 − cos Δt
q = lim
Δs → 0 (2πr )2 + L2 Δs
2 2 π⋅Tr Δt
= lim
Δs → 0 (2πr )2 + L2 2 Δs (10)

Take Eq. (5) into the Eq. (10), Then we get the theoretical so-
lution for clench compression:

→ 4π 2Tr
q =
(2πr )2 + L2 (11)

While in the above FE model, this clenching effect of each in-

Fig. 5. Force diagram of the infinitesimal Δs.
finitesimal element can be obtained as the resultant force of the X,
Y spring reactions. Therefore, to validate the accuracy of the FE
under longitudinal stretching. This clench effect is the other main model, four prestressed cable examples (two spiral strands, and
source of inter-wire extrusion. The magnitude of this effect is two semi-parallel wire cables) are built to calculate and compare
decided by the twisting angle and the wire winding radius Fig. 5. the clench forces from both FE simulations and theoretical calcu-
Given the wire helical pinch L and winding radius r, a spiral lations. Fig. 6 presents the cross section arrangement and locations
wire can be described as: of calculated wires (marked in red). For an inner layer wire, the
contact compression at its concave side is not only affected by the
⎧ x = r cos t
⎪ pinch effect of longitudinal stretching but also by the accumulated
⎨ y = r sin t t ∈ [0, π ]. compaction from outer layer wires. Then only outer layer wires are
⎪
⎩ z = L⋅t /2π (5) selected for comparison to exclude this accumulation effect.
Previous research [21] reported that friction exerts little effect
Where t is the rotation angle of a point at the helical wire. Given T
on longitudinal stretching performance; therefore, we simplify the
as the wire force, then vector force at each end of an infinitesimal
F–E relation of the Z springs to the bilinear springs with 10 N/mm
piece of helical wire can be computed as follows:
ultimate friction resistance. Wire diameters are 5 mm for all of the
⇀ 2π⋅T ⎛ L ⎞ cables, and an axial traction of 100 MPa (1963.5 N point force) is
T = ⎜ −r cos t , − r sin t , − ⎟.
(2πr )2 + L2 ⎝ 2π ⎠ (6) applied to each wire. Winding radius, pinch length and lay angle
for calculated wires are given in Table 1. Wire beam elements at
the midsection of cable are selected to exclude boundary effect.
⇀ 2π⋅T ⎛ L ⎞
T′ = ⎜ r cos ( t + Δt ), r sin ( t + Δt ), ⎟. Wire force T can be acquired from the FE model, and then the
(2πr )2 + L2 ⎝ 2π ⎠ (7) theoretical clench compactions can be calculated with Eq. (11). The
According to the static equilibrium, following principle can be comparison indicates that the clench forces derived from the FE
obtained: model agree well with the theoretical results, thereby partially
validating the effectiveness of the FE model. Nonetheless, the
⇀ ⇀ ⇀ spiral strand exerts a more obvious pinch effect than the semi-
q ⋅Δs = T + T ′ . (8)
parallel wire cable due to the larger lay angle and shorter pinch
Thus, the stress on this short piece of wire can be calculated length of the former.

Fig. 6. Wires selected for the clench force calculation and comparison. (For interpretation of the references to color in this figure, the reader is referred to the web version of
this article).
 Y. Yu et al. / International Journal of Mechanical Sciences 113 (2016) 196–210 201

Table 1
Comparison of the theoretical calculations and the FE model results.

Wire number Cable standard Winding radius Pitch length L(mm)/lay Element length Wire force (N) Clench force of the FE Calculation of Eq. (11)
(mm) angle(º) (mm) model (N)

I-1 Spiral strand 5 400/4.49 20.06 1966.7 47.29 48.38

II-1 Spiral strand 5 360/4.99 15.06 2016.20 44.85 45.89
II-2 Spiral strand 10 360/9.9 15.23 1963.50 86.58 88.37
III-1 Semi-parallel wire 5 1200/1.5 20.01 1967.20 3.72 5.39
cable
III-2 Semi-parallel wire 10 1200/3 20.03 1963.10 10.82 10.75
cable
III-3 Semi-parallel wire 8.66 1200/2.6 20.02 1964.50 9.60 9.31
cable
IV-1 Semi-parallel wire 15 1200/7.75 20.06 1964.00 16.61 16.10
cable
IV-2 Semi-parallel wire 13.23 1200/6.84 20.05 1961.20 14.83 14.19
cable

3.3. Determination of friction resistance for bending simulation during the bending process can be captured. The flow diagram for
the process of determining friction resistance is shown in Fig. 7.
Because that the FE model can consider the wire compaction The first step involves applying axial traction to obtain the in-
effect by itself, once the sheath pinch effect and helical wire itial friction resistance distribution of each Z spring. Then initially,
clenching effect are settled, the contact extrusion and corre- all the Z springs are assumed to have a same bilinear relation with
sponding friction resistance among the wires can then be de- a friction resistance of 10 N. Calculate the cable model under cir-
termined. In this study, the value of the Coulomb friction coeffi- cumferential compression and axial stretch, and then retrieve the
cient parameter μ is 0.5, and bending is ensured by the lateral internal force at X, Y springs to calculate contact extrusion and the
displacement load on the middle of the cable. This lateral deflec- ultimate friction resistance for corresponding Z spring. Subse-
tion is added by many small increments through load steps. Fric- quently, we repeat the axial stretch calculation using the updated
tion resistance at Z springs is determined through iterative cal- ultimate friction relations and recalculate the friction resistance
culations, which includes recalculating the contact friction re- again to derive a more reasonable friction value from the iteration.
sistance and updating the F–E relation at each load step. Therefore, As per a comparison, the obtained friction resistances of the first
the contact compression and ultimate friction resistance variations two iterations are already quite close. Therefore there only has one

Fig. 7. Flow chart of the friction determination and updating processes.

202 Y. Yu et al. / International Journal of Mechanical Sciences 113 (2016) 196–210

recalculation and refreshing process for each Δ increment of 4. Validation of the FE model and discussion on cable bending
flexural deflection.
During the bending process, the F–E relation of Z spring is 4.1. Comparison of semi-parallel wire cables
modified through adding points. Once the first two friction re-
determination procedures under tension condition are completed, Wu [22] once performed a series of bending experiments on
ultimate friction resistance at Z springs are denoted by structural cables, in which different cable styles (stranded ropes,
semi-parallel wire cables, and spiral strands), end conditions and
F0 (1) , F0 (2) , F0 (3) ... F0 (i ) ...which are utilized to form a bilinear
pretension conditions were all considered. In particular, the ex-
relation with the turning point (E0 (i ) , F0 (i )), as depicted in Plot 2 of
perimental data for Φ7 serial semi-parallel wire cables are com-
Fig. 7. Then apply Δ increment of lateral displacement at the
prehensive that can be used to validate the proposed advanced FE
midsection of cable, and retrieve X, Y spring forces ( Fx (i ), Fy (i )),
model. In Wu's study, all the cable specimens were 1050 mm long,
resultant contact extrusion, and the force and elongation at Z
and the distance between two supporting pads (bending span)
spring ( E (i )). Subsequently a series of comparisons and calcula- was 820 mm. Bending deflection was applied with a hydraulic jack
tions is performed to determine the additional point (EΔ (i ) , FΔ (i )). If at the mid-span (Fig. 8(a)); thus in the FE simulations, bending
the Z spring approaches the ideal plastic state (that is, the re- deflection was ensured with displacement loading at the center
trieved elongation goes over the previous limit ( E (i ) > E0 (i ), which node of the mid-span section. The test included three different end
also represents a steadily slipping state)), the new spot is calcu- conditions, namely free end, welded end and prestressed end to
lated as follows: explore the boundary effect, of which the corresponding experi-
mental images are depicted in Fig. 8 and Fig. 9.
⎧
⎪ EΔ (i ) = E (i )
The different end conditions can be realized through different
⎨ . node coupling conditions at cable end section: in the simulation of
⎪ 2 2
⎩ FΔ (i) = μ Fx (i) + Fy (i) (12) free-end test, rigid plane restraint equations are applied to the X, Y
displacement freedoms for the cable end nodes, while releasing Z
If the Z spring does not achieve the ideal plastic state, the ad- direction displacements (Z-axis runs along the longitudinal di-
ditional point is determined as: rection of the cable, as per Fig. 9(b)). With this setting, the cable
end behaves as the plane section along the X, Y directions while
⎧
⎪ F (i ) = μ F (i )2 + F (i )2
allows nodal displacement difference along Z direction; When
x y
⎨ Δ , restraint equations are applied to all three direction displacement
⎪
⎩ Δ
E (i ) = E 0 (i ) + FΔ (i ) − F0 (i) /e (13) freedoms, the entire section acts as a rigid plane section that ro-
tates collectively without permitting any Z displacement differ-
Where e corresponds to the elastic modules of steel. F–E relation of ence or slip. This setting is used to simulate the welded-end
the Z spring changes from bilinear to trilinear, as shown in Plot bending test, as in Fig. 9(d); The prestressed cable had its pre-
3 of Fig. 7. Subsequently, the Δ displacement bending calculation tension integrated through a steel block that welded to the cable
and the aforementioned comparison processes are repeated with at one end and pin connected to an actuator system at the other
the newly obtained trilinear Z spring relations to yield a refreshed end. Therefore in the FE a short beam is installed at each cable end
point (E′Δ (i ) , F ′Δ (i )), which is then settled as the ultimate Z spring model to simulate the steel block. The cable end nodes are con-
nected to the short beam with all degree rigid plane couplings to
F–E relation at the end of Δ displacement bending (Plot 4 in Fig. 7).
simulate the welding boundary. Point forces are applied at the
Next, 2Δ lateral displacement is applied and the aforemen-
other end of short beams to involve pretension and X, Y dis-
tioned process is repeated twice to determine (E′2Δ (i ) , F ′2Δ (i )).
placement constraints are set at the short beam ends and sup-
Hence, we can generate a multi-linear Z spring F–E relation with
porting pad locations (Fig. 9(f)).
points representing friction resistance capability at Δ, 2Δ, 3Δ…
Semi-parallel wire cables are generally wrapped with high-
bending states a several iteration calculations. Under nΔ flexural density polyethylene (HDPE) sheath to protect the inside wires
deflection bending, there will have nþ1 points at each F–D rela- from corrosion, and all the wire cable specimens kept the HDPE
tion (Plot 5 in Fig. 7). Once all these points are settled, the model is sheath during the bending tests. While in the numerical modeling,
then recalculated for the last time with nΔ bending deflection to the HDPE sheath part is hard to simulate due to the varying and
get the desired bending simulation with considering the contact unsettled material properties. Then the FE models are unable to
friction variation. determine the bending strength contribution from HDPE sheath

Fig. 8. Bending test setup and the advanced FE model.

 Y. Yu et al. / International Journal of Mechanical Sciences 113 (2016) 196–210 203

Fig. 9. Final details of the cable specimens and the corresponding FE model settings.

but can only consider the sheath compaction with centripetal stiffness with the corresponding free end ones, but presents a
forces. The thermal expansion coefficient is about 1.4
10  4 C  1, higher stiffness than that of free-end cables at large curvature
the elastic modulus is about 150 MPa and the temperature dif- state. The comparison indicates that the welded-end measure-
ference during the hot extrusion process is about 24 °C for HDPE ment can increase cable rigidity by restraining wire slip at the
sheath, then the compaction effect can then be estimated with cable ends, but this strengthening effect is limited and only ob-
Eqs. (2) and (3); corresponding parameters and calculations are vious in large size cables and at large curvature state.
listed in Table 2. The transition regions between the “elastic” and “hardening”
Figs. 10 and 11 present the applied transverse load versus the states in the bending tests are larger than those obtained with
sag deflection relation according to both test results and FEM si- FEM, and so does the flexural strength, which may originate
mulations, respectively. Fig. 10 illustrates the unstressed cable mainly from the bending contribution of HDPE sheath. None-
bending situations for free-end and welded-end cables. The FEM theless, the tangent stiffness of the FEM simulations during both
simulations all present a comparatively lower flexural strength the initial period and the subsequent stabilized state are con-
than that in the experiments, but had similarly varying trends and sistent with the experimental results. Given the similarly varying
relations for both free-end and welded-end conditions. The un- patterns, the FE models showed their effectiveness in analyzing
stressed cables exhibit nonlinear bending behavior that is analo- the bending behavior of unstressed cables.
gous to the performance of an elastic–plastic material beam with Fig. 11 displays the flexural strength relations of cables under
hardening in both test results and FEM simulations. The curves different pretension levels. The FE modeling results exhibit in-
start with a large tangent module which indicates a high initial creasing trends of flexural strength and tangent stiffness that are
flexural stiffness during the beginning period. Then as the curva- similar to those observed in the experimental data. In the test
ture increases, the tangent stiffness decreases and gradually shifts results, the strength magnitude and tangent stiffness are increas-
to a linear relation in the latter loading period. The effective ingly related to pretension level, while the FE modeling results do
bending stiffness of welded end cables all shows a similar initial not indicate a significant strengthening effect from the axial
204 Y. Yu et al. / International Journal of Mechanical Sciences 113 (2016) 196–210

Table 2
Parameters of the calculated cable models.

Cable type Diameter of wire Thickness of HDPE Pitch length L Simplified cen- Cable section Testing scheme
bundle (mm) sheath(mm) (mm) /Lay angle tripetal force(N)
(º)

37Φ7 Semi-parallel 49 8 1800/4.2 28.15 Free-end and welded-end free

wire cable bending;Bending with 5, 8, and 10
Ton pretensions

55Φ7 Semi-paral- 58 9.5 2182/4 25.07

lel wire cable

61Φ7 Semi-parallel 63 9.5 2400/4.2 25.07

wire cable

traction. Due to the decreased strength contribution of HDPE magnitude and is even smaller than the extrusion variation during
sheath, the gap between the experimental curves shrinks as the bending. Therefore, the increasing effect of pretension on bending
increasing of cable dimension. Under the same stretching force, strength remains weak.
the axial stress decreases in large cables, and so does the axial
stress in HDPE sheath. Therefore, flexural strength exerts weak 4.2. Validation of the spiral strand model
influences on large cables, thus reducing the difference among the
force–deflection relations. Considering the interference of HDPE Wu [22] also conducted free-end bending tests on two Galfan
sheath, the performance of this FE model in terms of analyzing the spiral strands (Φ38 (1
37) and Φ48 (1
91)), as represented by
bending behavior of tensioned cables is promising. the experimental specimens shown in Fig. 12. Galfan strand is a
The FEM results also predicted the aforementioned elastic–plastic new type of structural cable product that has advanced anti-cor-
behavior during the initial loading stage, which phenomenon was rosion coating at wire surfaces, which ensure an effectively cor-
hard to be captured in the test. Pretension mainly affected the rosion resisting ability on its own and release the cable from ex-
transition strength, with its value get increased with the magnitude ternal HDPE sheets. Wires in the Galfan strand are alternatively
of axial traction. Hong [1,23] investigated cable bending mechanisms twisted between adjacent layers, which exhibits the same wire
and noted that all wires were initially in a stick state; when the cable winding pattern as the aforementioned spiral strand (Fig. 12(a),
was bent, the imbalanced tension force developed in each cable (c), and (d)).
segment tended to slide wires relative to its neighbors. The overall Given the alternative twisting patterns of the spiral strand, an
bending strength originated from two parts: the flexural stiffness of individual wire will have point contact with adjacent layer wires
the wire bundle, and the composite effect from inter-wire friction and have line contact with its neighbors in the same layer. This
resistance. The axial stretch mainly affects the clench effect in ten- contact mode can be simplified into a four-point contact set, which
sioned spiral wires and the corresponding friction resistance, and includes two contact spots for adjacent wires in the same layer
then indirectly influences the cable bending behavior. Therefore, the and two other single-point contact spots for the inner and outer
initial contact friction is large in high tension conditions and reflects layer wires. Therefore, the wire section can be replaced by four
a strong capability to resist wire slips in the cable. However, this rigid beams that connect the four contact points and wire center,
friction increase is significantly lower than the overall friction as with the six rigid beam sets in the semi-parallel wire cable

Fig. 10. Bending behavior of unstressed cables.

 Y. Yu et al. / International Journal of Mechanical Sciences 113 (2016) 196–210 205

Fig. 11. Bending behavior of prestressed cables.

(Fig. 12(b)). The spiral specimens in the tests had steel wire clamps cross section, each layer only has six wires point contacted with
installed at two cable ends (Fig. 12(e)). Moreover, test results the inner or outer layer wires, with the rest wires spanning over
showed that the clamps generated a strong bonding effect at the the adjacent layers and having point contacts at different cross
cable ends, since the strand ends gave a rigid plane rotation during sections along the cable. At Section 1, all of the contact spots are
significant bent deflection. Hence, rigid plane restraint equations supposedly located at the hexagon diagonals (the red spots in
are applied to the end section nodes on three displacement free- Section 1). Δ is the interval of each cable section, and the twisting
doms to simulate the clamping effect or the rigid plane rotation angles for each wire layer are θ2, θ3, θ4, θ5, and θ6 from inside to
conditions. the outside. Those four rigid beam sets are also marked and
During the real fabrication of spiral stand, lay angle of the he- numbered by location for clear presentation. For instance, the label
lical wires is controlled within a range of 10–15°. Moreover, due to “wire (3, 4)” indicates that this wire is located in the third layer and
the alternative twisting in Glanfan strand, the wire construction is the fourth wire along anti-clockwise direction. Therefore in
differs significantly from that of a semi-parallel wire cable. The Section 1, Wire (5,9) is point contacted with Wire (6,11), and Wire
principle idea of the proposed model is to divide the helical wires (3,3) is next to Wire (4,4). As the helical wires spanning from
into many short pieces and building spring models at those con- Sections 1 and 2, each layer shifts around the cable axis with θ2, θ3,
tact locations. Then the wire contact locations and their varying θ4, θ5, and θ6 angles, and then the contact pattern get changed. In
patterns need to be understood at first. Section 2, Wire (5,9) reaches Wire (6,10) while Wire (3,3) makes
Fig. 13 illustrates a short piece from a Galfan 48-strand FE contact with Wire (4,3). While in Section 3, wires (5, 9) and (3, 3)
model with three cable sections, in which the wire arrangements are point contacted with Wire (6, 9) and Wire (4, 2), respectively.
and contact locations are depicted as well. In any selected cable Following this pattern, if an individual wire (wire (i, j)) has point

Fig. 12. Experimental specimens and FE model of the Galfan cable.

206 Y. Yu et al. / International Journal of Mechanical Sciences 113 (2016) 196–210

Fig. 13. Cable section settings and determination of the contact location.

Table 3
Parameters of the calculated Galfan Strand.

Cable type Diameter of wire bun- Wire θ1(º)/ Pitch length L θ 2(º) / Pitch length L θ 3(º) / Pitch length L θ4 (º) / Pitch length θ5(º) / Pitch length
dle (mm) number (mm) (mm) (mm) L(mm) L(mm)

Glanfan38 5.6 37 11.78/170.5 13.62/294.5 16.68/360 – –

Glanfan48 4.4 91 10.15/155.7 12.60/250.4 13.96/338.82 14.23/443.1 15.04/523.6

1500
Galfan 38 (37 wire) Galfan 48 (91 wire)
600
Lateral force (N)

Lateral force (N)

1000
400

500
200
FE modelling FE modelling
Test result Test result
0 0
0 1 2 3 4 5 0 1 2 3 4 5
Bending deflection (mm) Bending deflection (mm)
Fig. 14. Validation of the FE modeling method for the Galfan spiral strand.

contact with an outer layer wire (wire (iþ1, k)), then after one Δ be determined to ensure the coincident node pair assumption.
span twisting this wire will shift to wire (iþ1,k þ1) or wire (iþ1, Table 3 gives the calculated parameters for the Φ38 and Φ48
k  1), depending on the winding direction. This rotating pattern Galfan strand.
can be described as follows: Fig. 14 compares the applied transverse load versus sag de-
flection relation of the Galfan strands from both test data and FE
360
θi + θi + 1 = simulations. The two curves are pretty consistence with each
6i (14)
other, which supports the effectiveness of this simplified FE
Where θi and θi + 1 are the twisting angle increments of wire layers i method on spiral strand simulations. The Galfan strand also ex-
and iþ 1, respectively, and 360/6i is the central angle occupied hibits a bilinear elastic–plastic-like flexural strength–deflection
by a single wire in layer iþ1. With this relation, a reasonable response, of which the “plasticity development” region represents
length of cable piece Δ and the twisting angle θ for each layer can the wires stick-to-slip state transition during bending.
 Y. Yu et al. / International Journal of Mechanical Sciences 113 (2016) 196–210 207

Fig. 15. Bending deformation under free-end bending (55-wire semi-parallel wire cable; unit: mm).

Fig. 16. Bending deformation under welded-end bending (55-wire semi-parallel wire cable; unit: mm).

Fig. 17. Bending deformation under 8-ton pretension (55-wire semi-parallel wire cable; unit: mm).

The initial bending stiffness of the FE result is slightly lower conditions cannot be completely replaced by Coulomb friction. In
than that of the test curve; by contrast, the stabilized tangent addition, the Coulomb friction coefficient is 0.5 for all the contact
stiffness observed during a significant bending period (post-tran- spots, which is an assumed and optimized value, while the real
sition region, when the deflection is larger than 2 mm) is higher friction coefficient is unclear and uneasy to get (Different friction
than that of the test result. The difference is mainly attributed to coefficients were tried before this study and found that Coulomb
the simplicity of inter-wire friction: in the actual strand, the anti- friction 0.5 could ensure a good prediction of cable stiffness for
corrosion coating and fabrication process may generate a complex both semi-parallel wire cables and Galfan strands). Considering
friction condition or mechanical bonding among the wires which these simplifications, the advanced FE models can still predict
208 Y. Yu et al. / International Journal of Mechanical Sciences 113 (2016) 196–210

Fig. 18. Wire force distribution of a 55-wire semi-parallel wire cable under different conditions.

Fig. 19. Deformation and axial force distribution of the Galfan 48 strand.

similar flexural behaviors for both semi-parallel wire cables and ends (free-end bending condition, Fig. 15). But for welded end and
spiral strands, indicating the effectiveness of the proposed FE prestressed cables, terminations or clamps at both cable ends can
model. prevent the wires from slippage or micro-rotation, and then the
uneven wire forces and the induced additional moment may cause
4.3. Discussion on cable bending behaviors sinusoidal deformation to compensate the unwinding trend and
wire slips (Fig. 16).
Due to the concurrent winding in semi-parallel wire cables, the Previous theoretical studies always take the plane-section as-
applied transverse load will produce an uneven force system, and sumption and one direction bending as the basic principles for
induce an unwinding trend in helical wires, as well as the frictional bent cable, making it hard to consider the torsional effect and
slip. Without end constraints, this unwinding trend can be released combined bending behaviors in cable bending. While during the
through wire slips or micro rotational displacements at the cable test, slight out of bending plane deflection was observed. And
 Y. Yu et al. / International Journal of Mechanical Sciences 113 (2016) 196–210 209

because that the propose FE model is a 3D analyzing method, it hardening. Welded end can add cable bending stiffness during
has the ability to capture this combined bending process and large curvature bending states. Pretension force can increase
predict the uneven tension system. Figs. 15–17 show the de- the transition strength, but this strengthening effect is limited.
formation conditions of semi-parallel wire cables under different (2) During the bending of semi-parallel wire cables, the concurrent
conditions (given the 55-wire cable as a representative). All of the twisting wires produces an unwinding trend and sinusoidal
results are retrieved under the 4 mm bending deflection condition, deformation out of bending plane. The magnitude of this si-
and the deformations are scaled 15 times for displaying. With the nusoidal deflection is influenced by the end conditions, that a
parabola-like deflection along the bending plane (along the X di- strong end clamping effect can facilitate the spread of the
rection), sinusoidal deformations are initiated along the perpen- bending stress gradient and aggravate sinusoidal deformation.
dicular direction (along the Y direction). In Fig. 15(b), the free end (3) Spiral strand has alternative twisting permutation between
allows wire slip and releases the unwinding trend from cable ends, adjacent layers, which has a canceling effect on the torsional
thus forming an inclined section and a comparatively small sinu- effect and unwinding trend. Galfan strand has negligible si-
soidal deflection along Y direction. In welded-end cable (Fig. 16), nusoidal deformation during bending and the flexural pattern
the wire ends are coupled but can still rotate together. While for is similar to that of a simple supported beam.
prestressed cable no end displacement and section rotation are
allowed due to the clamped boundary from the prestressing Since the real mechanism of friction is not quite clear and
measurement in the test. Then the unwinding trends are getting certain simplifications are adopted in the inter-wire force calcu-
harder to be released, and more obvious sinusoidal deformations lation, the FE method still requires further improvements and
will be observed from free end cases to welded end and pre- modifications to enhance simulation accuracy. Nonetheless, this
stressed end conditions. Moreover, the welded-end cable exhibits study is the first to use the FE model to simulate long and large
a deflection pattern between the simple beam bending and dimension cables (820 mm long, composed of a maximum of 91
clamped beam bending behaviors. wires in this study) in a short time. Although the advanced FE
Fig. 18 shows the axial wire force distribution of 55-wire semi- model is mainly applied to cables that are subjected to tension and
parallel wire cables with 4 mm deflection. Legends for the free- point load bending conditions in this work, it can also be used to
end and welded-end conditions are modified and wire force at simulate wire degradation mechanisms, such as corrosion and
mid-span section and quarter-span section are also displayed for wire breaks, given its efficiency and easy manipulation.
better comparison. Helical wires exhibit a gradient tension dis-
tribution for all the three cases but the ranges of this gradient are
different. Welded end measurement can add integrality to the Acknowledgments
cable, and then lead to a higher extent and wider range of bending
stress than the free end case. While for the tensioned cable, due to This work is sponsored by the National Natural Science Foun-
the strongly clamped ends, the uneven force system cannot be dation of China (Grant no. 51478310).
released from end section rotation or accommodated through in-
ter-wire slippage either. Therefore, the high tension gradient
spreads along the entire cable range, as shown in Fig. 18(c).
References
Fig. 19 illustrates the bending deformation and wire force dis-
tribution of the Galfan 48 strand under 3 mm bending deflection.
[1] Hong K, Der Kiureghian A, Sackman J. Bending behavior of helically wrapped
Due to the alternative twist between adjacent layers, the induced cables. J Eng Mech 2005;131(5):500–11.
unwinding effect from tensioned helical wires is canceled out, [2] Giglio M, Manes A. Life prediction of a wire rope subjected to axial and
resulting negligible aforementioned sinusoidal deformation and bending loads. Eng Fail Anal 2005;12(4):549–68.
[3] Raoof M, Hobbs RE. The bending of spiral strand and armored cables close to
similar deflection pattern as simple supported beam. Galfan strand terminations. J Energy Resour Technol 1984;106(3):349–55.
has different wire force distribution from semi-parallel wire cable, [4] Raoof M. Interwire contact forces and the static, hysteretic and fatigue prop-
that high tension or compression distributes in inner layers in- erties of multi-layer structural strands. London: University of London; 1993.
[5] Durelli AJ, Machida S, Parks VJ. Strains Displacements on a Steel Wire Strand
stead of the outmost layer (Fig. 19(d)).
1972;84(6):85–93.
[6] Papailiou KO. On the bending stiffness of transmission line conductors. ETATS-
UNIS.New York, NY: Institute of Electrical and Electronics Engineers; 1997.
5. Conclusions [7] Filiatrault A, Christopher S. Flexural properties of flexible conductors inter-
connecting electrical substation equipment. J Struct Eng 2005;131(1):151–9.
[8] Poffenberger JC, Swart RL. Differential displacement and dynamic conductor
An advanced FE modeling method is proposed and formulated for strain. IEEE Trans Power Appar Syst 1965;84(4):281–9.
structural cables on bending and wire sliding problems. Cable com- [9] Claren R, Diana G. Dynamic strain distribution on loaded stranded cables. IEEE
Trans Power Appar Syst 1969;PAS-88(11):1678–90.
ponents and their influencing factors like twisted wires, contact [10] Lanteigne J. Theoretical estimation of the response of helically armored cables
conditions, friction resistance and slip are all specially considered: to tension, torsion, and bending. J Appl Mech 1985;52(2):423–32.
Spiral wires are replaced with helically arranged short-beam ele- [11] Judge R, Yang Z, Jones SW, Beattie G. Full 3D finite element modelling of spiral
strand cables. Constr Build Mater 2012;35(0):452–9.
ments; Wire section and contact spots are simulated using radially
[12] Jiang W-G. A concise finite element model for pure bending analysis of simple
placed rigid beam elements; Contact compression and the resultant wire strand. Int J Mech Sci 2012;54(1):69–73.
friction resistance are calculated and settled using the spring ele- [13] Montoya A, Waisman H, Betti R. A simplified contact-friction methodology for
ments positioned along different directions. And then the compres- modeling wire breaks in parallel wire strands. Comput Struct 2012;100–101
(0):39–53.
sion force, friction calculation, and contact pattern are all considered [14] Yu Y, Chen Z, Liu H, Wang X. Finite element study of behavior and interface
and determined for both semi-parallel wire cables and spiral strands. force conditions of seven-wire strand under axial and lateral loading. Constr
The proposed FE model is validated through experimental data, and Build Mater 2014;66(0):10–8.
[15] Usabiaga H, Pagalday JM. Analytical procedure for modelling recursively and
then employed to explore the deflecting pattern and wire force
wire by wire stranded ropes subjected to traction and torsion loads. Int J Solids
distribution of two structural cables. Following conclusions can be Struct 2008;45(21):5503–20.
drawn regarding cable bending behavior: [16] Zhou W, Tian H-Q. A novel finite element model for single-layered wire
strand. J Cent South Univ 2013;20(6):1767–71.
[17] Gopinath D, Shibu G, Thanigaiyarasu G. Theoretical Estimation of Stiffness of
(1) The bending behavior of semi-parallel wire cables is nonlinear Stranded Cable Subjected to Constrained Bending.
and resembles that of an elastic–plastic metallic beam with [18] Papailiou KO. Bending of helically twisted cables under variable bending
210 Y. Yu et al. / International Journal of Mechanical Sciences 113 (2016) 196–210

stiffness due to internal friction, tensile force and cable. Switzerland; 1995. linear analytical models for steel wire strands under axial loading, using a 3D
[19] Sun G, Chen Z, Liu Z. Analytical and experimental investigation of thermal FE model. Int J Mech Sci 2007;49(11):1251–61.
expansion mechanism of steel cables. J Mater Civil Eng 2011;23(7):1017–27. [22] Wu X. Research on bending performance of cable tianjin.China: Tianjin Uni-
[20] Yu Y, Chen Z, Wang X. Refined simplified finite element model of cable and its versity; 2013.
sensitivity analysis. J Tianjin Univ Sci Technol 2015;48(Suppl. 1):96–101. [23] Hong K. Dynamic interaction in cable-connected equipment.Berkeley: Uni-
[21] Ghoreishi SR, Messager T, Cartraud P, Davies P. Validity and limitations of versity of California; 2003.