Vous êtes sur la page 1sur 13

International Journal of Mechanical Sciences 115-116 (2016) 202214

Contents lists available at ScienceDirect

International Journal of Mechanical Sciences


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

Analytical and nite element modelling of the elasticplastic behaviour


of metallic strands under axialtorsional loads
Francesco Foti a,n, Alessandro de Luca di Roseto b
a
Politecnico di Milano, Department of Civil and Environmental Engineering, P.zza L. da Vinci 32, Milano 20133, Italy
b
Loughborough University, School of Civil and Building Engineering, Sir Frank Gibb Building, Loughborough LE11 3TU, England, United Kingdom

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

Article history: In this work a new formulation for modelling the elasticplastic behaviour of metallic strands subjected
Received 18 February 2016 to axialtorsional loads is presented. Simple and accurate cross sectional constitutive equations are de-
Received in revised form rived, fully accounting for the evolution of plastic deformations in the wires, starting from a description
20 June 2016
of the internal structure of the strand. The proposed approach is suitable both for straightforward
Accepted 21 June 2016
analytical calculations as well as for implementation into nite elements for the large-scale structural
Available online 22 June 2016
analyses of cable structures. A full three-dimensional (3D) nite element (FE) model, based on a para-
Keywords: metric description of the strand internal geometry, is also developed. The results of both the FE model
Strands and the analytical formulation are validated with reference to a well-documented physical testing
Axialtorsional loads
campaign and a well-established linearly elastic literature model. Additional analyses are then performed
Elasticplastic behaviour
to carefully assess the validity of the proposed mechanical formulation, for a wide range of strand
Finite element method
Curved thin rod theory construction parameters, by means of systematic comparisons against the results of the 3D FE model and
of a recent linearly elastic literature model.
& 2016 Elsevier Ltd. All rights reserved.

1. Introduction strands under a combination of axial force and torsional moment


(detailed reviews can be found e.g. in [3,10]). Most of them rely on
Metallic strands are made of helical wires, twisted around a the assumption of linearly elastic wires, thus providing an insight
straight core and grouped in concentric layers. The core is typically into the behaviour of strands under service loads. Within this
another wire (core wire), which contributes to sustain the external context, analytical formulations (see e.g. [4]) have been developed
loads and provides a radial support to the layers. The simplest to estimate both the stress state of wires as well as global response
construction consists of a single layer surrounding the core. Six parameters, such as the coupled axialtorsional cross sectional
outer wires are usually employed in this basic case, which will be stiffness matrix. Only few works, instead, considered the evolution
referred in the following as simple strand. A review of the most of plastic deformations under axialtorsional loads. Jiang et al. [13
common strand typologies, including details on their internal 15] proposed a rened nite element (FE) approach, which takes
into account plastic deformations and allow for an accurate de-
geometry, mechanical properties and manufacturing process, can
scription of internal contact conditions. More recently, procedures
be found e.g. in [9].
for developing full three-dimensional (3D) elasticplastic FE
Due to their ability in carrying large axial forces with relatively
models of metallic strands have been proposed e.g. by Judge et al.
small dead-loads, metallic strands are efcient structural mem-
[17] and by Yu et al. [30], while Imrak and Erdnmez [12] adopted
bers, widely employed in mechanical and civil engineering appli-
the FE method to study wire ropes with complex cross sections
cations. Moreover, they can be helically wound to form wire ropes, under the assumption of elasticplastic material behaviour. Due to
which are used e.g. in hoisting devices, tethered marine structures, their huge computational cost, however, rich FE models cannot be
and suspended bridges. A distinctive feature of strands is the successfully applied for simple calculations or large-scale struc-
coupling between the axial and torsional behaviour, due to the tural analyses, which are typical of engineering applications. As a
helicoidal geometry of the wires. consequence, the post-elastic behaviour of cable structures is ty-
Many studies have been devoted to investigate the response of pically investigated by considering simple phenomenological uni-
axial constitutive laws (e.g. [16,18,19]), which do not take into
n
Corresponding author.
account the peculiar internal structure of strands and the direct
E-mail addresses: francesco.foti@polimi.it (F. Foti), consequences on the mechanical response, such as axialtorsional
a.delucadiroseto@lboro.ac.uk (A. de Luca di Roseto). coupling.

http://dx.doi.org/10.1016/j.ijmecsci.2016.06.016
0020-7403/& 2016 Elsevier Ltd. All rights reserved.
F. Foti, A. de Luca di Roseto / International Journal of Mechanical Sciences 115-116 (2016) 202214 203

To overcome these limitations, a new approach is presented in The symbol in (1) denotes the swept angle, i.e. the angle
this work to model the elasticplastic behaviour of simple strands which the projection of the position vector on the plane x1 = 0
subjected to axialtorsional loads. Cross sectional constitutive denes with the axis x2. The subscript 0 is adopted to identify the
equations are derived starting from a description of the internal value of the swept angle at x1 = 0.
structure of the strand, herein considered as a composite struc- The orientation of the wire cross sections, then, can be de-
tural element and fully accounting for plasticity effects. The aim is scribed by specifying a local system of axes, attached to the heli-
to provide for a sound mechanical framework, suitable for both coidal centreline. To this aim, the right-handed SerretFrenet unit
simple analytical calculations as well as implementation into nite { }( )
vectors fi( ) i = 1, 2, 3 are introduced, such that f1( ) is the
elements for the large-scale structural analyses. Each wire of the tangent vector, while f2( ) and f3( ) are, respectively, the normal
strand is individually modelled as a curved thin rod in the fra- and binormal unit vectors of the wire centreline. The SerretFrenet
mework of the classic KirchhoffClebschLove theory [25]. The unit vectors can be evaluated starting from (1) (see e.g. [21]) and
Von Mises yield criterion is adopted to model the interaction be- related to the unit vectors { ei} by means of a rotation tensor,
tween normal and tangential stresses on the wire cross sections,
w ( ), i.e. fi( ) = w ( )ei , ( i = 1, 2, 3). By denoting as the lay
together with the well-known PrandtlReuss associated ow rule.
angle of the wire, i.e. the constant angle which the tangent vector
Kinematic equations are introduced to relate the axial strain and
f1 denes with the strand axis x1, the components of w ( ) with
the torsional curvature of the strand to the generalized strains of
the wires. Then, starting from the knowledge of the normal and respect to the basis { e } (here denoted as ) can be expressed
i w, ij

tangential stress distributions over the cross sections of the wires, as follows:
the resultant axial force and torsional moment of the strand are
cos( ) 0 sin( )
evaluated with equilibrium considerations.
The proposed mechanical formulation is rst directly assessed [ w, ij] = sin( )sin( ) cos( ) cos( )sin( )

with experimental results from the literature, secondly, with the sin( )cos( ) sin( ) cos( )cos( ) (2)
development of a full 3D FE model of the strand, which allowed us
to carefully investigate the performance of the proposed for- Starting from Eq. (1), the initial curvature ( ) and torsion ( ) of
mulation for a wide range of strand construction parameters. the wire centreline, which are of special importance in their me-
The results of both the analytical as well as the FE model are sin2( )
chanical modelling, can be dened, respectively, as = R and
preliminary validated with reference to a well-documented sin( )cos( )
benchmark. Comparisons are carried out with respect to both = R
(see e.g. [21]). It is worth noting that the lay angle can
available experimental results [27] and a well-established linearly be related to the helix radius and pitch through the simple geo-
elastic literature model [4]. Additional analyses are then per-
formed, for different strand constructions, to check the accuracy of
metric relation = tan1 ( ). Hence, the geometry of the external
2R
P
wires can be completely dened by specifying two construction
the proposed elasticplastic analytical formulation by means of parameters only, namely, the helix radius R and the pitch P (or
systematic comparisons against the results of the 3D FE model and equivalently the lay angle ).
of a recent linearly elastic literature model. Two basic internal contact modes can be distinguished (see
also [3]): radial contact (Fig. 2(a)) and lateral (or circumferential)
contact (Fig. 2(b)). In the rst case the external wires are in contact
2. Geometry of the strand with the core, but not among them. Accordingly, the helix radius R
is simply given by the sum of the diameters of the core and of the
A simple strand made of seven wires with circular cross section
is considered in this work, as depicted in Fig. 1. The geometry of
external wires, i.e. R =
1
2 (d 0 )
+ d . In the lateral contact case, in-
the internal structure is described with reference to the straight stead, the external wires are in contact with their neighbours, but
conguration of the strand. To this aim, a right-handed Cartesian not with the core wire. The helix radius, hence, turns out to be
system, with axes { x i} and unit vectors { ei} ( i = 1, 2, 3), is dened independent of the wire core diameter and can be evaluated as
d 3
such that x1 coincides with the strand centreline. A generic wire, R= 1+ , by assuming that wire cross sections are ellip-
2 cos2( )
then, is represented as a curved thin rod, by specifying for each
tical in a plane normal to the strand centreline [4].
cross section the position of the centroid and the orientation with
It is worth observing, however, that typical strand construc-
respect to the axes { x i}.
tions are characterized by clearances among the external wires, in
The centreline of the external wires is described through cir-
order to reduce interwire frictional effects and secondary tensile
cular helices, with radius R and pitch P, by means of position
stresses which can arise whenever the strand is bent [4,9]. As a
vectors with the following form:
consequence, a purely radial contact mode is assumed in this work
P to dene a reference geometric framework for the mechanical
x w( ) = ( 0)e1 + Rcos( )e2 + Rsin( )e3
2 (1) modelling of the strand, with both analytical as well as FE

Fig. 1. Geometry of the strand. (a) Side view. (b) Cross section. The effect of the lay angle on the projection of the cross sections of the wires is neglected in (b).
204 F. Foti, A. de Luca di Roseto / International Journal of Mechanical Sciences 115-116 (2016) 202214

Fig. 2. Cross section of the strand. Denition of contact modes: (a) radial contact and (b) lateral contact.

Fig. 3. (a) Straight strand subjected to axialtorsional loads. (b) Generalized stresses on the wire cross section. (c) Normal ( ) and tangential ( ) stress components on the
wire cross section.

techniques. This hypothesis amounts to consider: (a) a core wire 3. The elasticplastic mechanical model
with greater diameter than the external wires and (b) lay angles
smaller than a maximum value, max , corresponding to the onset of Let us consider a strand free from constraints, straight in the
lateral contact. The lay angle max can be easily evaluated by im- reference (undeformed) conguration and subjected to constant
posing that the helix radius R simultaneously satises the geo- axial force, Fs , and torsional moment, Ms (see Fig. 3(a)). Due to the
metric conditions for the radial and lateral contact previously in- symmetry with respect to the strand centreline (axis x1) and the
troduced. After some straightforward calculations, the following imposed boundary conditions, the element undergoes constant
geometrical condition can be obtained: axial strain, s , and torsional curvature, s . These generalized strain
variables are conjugated to the cross sectional stress resultants, Fs
and Ms , in the expression of the internal work per unit length of
the strand. A generalized axialtorsional constitutive law for the
3 d0
< max = arccos , with >1 strand cross section, hence, can be formally stated through the
2
d0 d
1 + 1
( ) ( )
functions Fs = Fs s , s and Ms = Ms s , s . The latter are derived in
d (3) this section starting from a description of the mechanical beha-
viour of wires, herein regarded as curved thin rods made of elas-
ticplastic material. Kinematic equations are introduced to relate
F. Foti, A. de Luca di Roseto / International Journal of Mechanical Sciences 115-116 (2016) 202214 205

the strains of the wires to the generalized strain variables of the and the generalized strain variables are constant along their
strand and a procedure for the evaluation of the stress distribution length. It can be concluded, hence, that the proposed kinematic
over the wire cross section is outlined. Then, the strand cross model satises the symmetry with respect to the strand centreline
sectional stress resultants Fs and Ms are dened through equili- (axis x1), known to characterize the mechanical problem under
brium considerations. study.
The wires can be modelled within the framework of the Focusing now on the wire cross section and recalling the hy-
KirchhoffClebschLove theory for curved thin rods [25]. Accord- pothesis of small strains, the normal ( ) and tangential ( ) strain
ingly, shear deformability is neglected and the cross sections are components at a generic point, identied by the polar coordinates
assumed to remain plane and normal to the wire centreline. Under ( r, ) dened in Fig. 3(c), can be expressed as:
the further assumption of small displacements and strains, the
(r , ) = w r cos()w3 ; (r , ) = rw1 (6a, b)
mechanical response of the wires can be studied within a local
reference frame, attached to the helicoidal centreline and with In order to introduce an elasticplastic constitutive law, the
axes directed as the SerretFrenet unit vectors { fi} introduced in strain components and can be additively decomposed into
Section 2 [7,8]. Within this context, the generalized stresses of the purely elastic ( e , e ) and plastic (p, p) contributions, such that
curved thin rod model are the axial force, Fw1, and the moments = e + p and = e + p . Accordingly, the normal ( ) and tan-
acting with respect to the directions of the vectors { fi}, i.e. the
gential stresses ( ) depicted in Fig. 3(c) are related to the elastic
torsional moment Mw1, and the bending moments Mw2 and Mw3 strain components only. Under the assumption of homogeneity
(see Fig. 3(b)). The work-conjugated strain measures are the wire
and isotropy of material and denoting, respectively, as E and the
axial strain w (i.e. the elongation of the wire centreline) and the
Young modulus and the Poisson coefcient, the following equa-
( )
mechanical curvatures wi i = 1, 2, 3 with respect to the axes of tions are introduced:
the local, wire-attached, reference system. The mechanical cur-
vatures are here introduced according to the denition provided in E
= Ee ; =
[8] and reported in Appendix A (see Eq. (A.1)). 2( 1 + ) e (7a, b)
The same mechanical model is also adopted to describe the
Consistently with numerical and experimental literature stu-
behaviour of the straight core, with minor modications: in the
dies on the axialtorsional behaviour of simple strands made of
case of the core, indeed, the unit vectors { fi} of the local reference
cold-drawn steel components [13,17,27]: (a) the elastic domain
frame simply coincide with those of the strand reference system
(herein denoted through the function F ( , )) is modelled by
{ ei}.
adopting the Von Mises yield criterion (Eq. (8a)), along with an
The axial strain of the wires, w , can be evaluated by exploiting
classic literature results. In fact, several authors (e.g. [20,23,4]) isotropic linear hardening law (Eq. (8b)) and (b) the evolution of
have shown that, during the axialtorsional loading of the strand: plastic strains is described though the classic PrandtlReuss asso-
(a) the core is subjected to the same elongation of the strand and ciated ow rule (Eq. (9)), herein introduced in rate form by de-
(b) the external wires undergo a transformation which preserves noting with a dot the derivative with respect to a time-like vari-
the geometrical shape of their centreline. Accordingly, the de- able:
formed centreline of the external wires is a circular helix, possibly
F ( , ) = 2 + 3 2 y( ) 0
characterized by different radius and pitch with respect to the
reference conguration. Variations of the helix radius can be due E
y( ) = y0 +
to: (1) the contraction of the diameters of both the external and E
1
E (8a, b)
the core wires due to the Poisson effect and (2) the deformation of
the internal contact surfaces between wires and core (wire at-
F
tening). These phenomena can signicantly inuence the response
of large-diameter strands, but can be practically neglected for the p =

very common case of simple strands with a steel core [26]. Starting F
=
from this latter hypothesis, Lanteigne [23] derived the following p (9)
kinematic equation:
The symbols y0 and E denote, respectively, the rst-yielding
w = cos2( )s + Rsin( )cos( )s (4) stress and the post-yielding Young modulus of the material (see
the bilinear uni-axial stressstrain curve in Fig. 4), while is an
( )
The mechanical curvatures wi i = 1, 2, 3 can be evaluated, internal variable which satises Prager's consistency conditions
under the assumption that the wire cross section rigidly rotates (see e.g. [24]):
with the cross section of the strand, by tailoring to the case at
study the general kinematics equations rst derived in [8]. Skip-
ping the calculations, which are fully reported in Appendix A, the
following expressions are obtained:

w1 = cos2( )s ; w2 = 0; w3 = sin( )cos( )s (5a, b, c)

By considering a lay angle equal to zero, Eqs. (4) and (5a, b, c)


hold true for the core wire also, which thus undergoes the same
combination of axial strain and torsional curvature as per the
strand element. On the other hand, due to their initial curved
conguration, the external wires are in general subjected to a
more complex strain state, characterized by a combination of axial
elongation, torsion and bending about the binormal unit vector of
the SerretFrenet triad ( f3). It is also worth noting that Eqs. (4) and
(5a, b, c) do not depend on the swept angle , introduced in
Section 2. As a consequence, all external wires behave identically Fig. 4. Bi-linear elasticplastic stressstrain curve of the material.
206 F. Foti, A. de Luca di Roseto / International Journal of Mechanical Sciences 115-116 (2016) 202214

0, F = F = 0 (10) dened in Eqs. (11a, b, c, d), by means of a standard numerical


integration technique.
The resultant axial force, Fw1, and moments, Mwi ( i = 1, 2, 3), of
the wire (see also Fig. 3(b)) are then evaluated through integration
over the cross sectional area A w :
4. The nite element model
Fw1 = A dA w ; Mw1 = A rdA w
(11a, b) A full three-dimensional (3D) nite element (FE) model has
w w
been developed using the ANSYS 15.0 software to simulate the
axialtorsional loading of straight strands. The aim is to provide an
Mw2 = A rsin()dA w ; Mw3 = A r cos()dA w
(11c, d) accurate numerical tool for a comparative assessment of the me-
w w

chanical formulation presented in Section 3.


Due to the already mentioned symmetry of the mechanical
The FE approach herein proposed relies on the parametric de-
model with respect to the strand centreline, all external wires are
scription of the internal geometry provided in Section 1, which
characterized by the same cross sectional stress resultants (11).
allows us to easily account for variations of the strand construction
This leads to a signicant simplication in the evaluation of the
parameters. A solid model of the strand is rst created in Solid-
strand cross sectional resultants, which, fully accounting for the
Works 2013 CAD, by extruding the round cross sections along the
projection from the reference system of the wire to the one of the
helix wire centrelines dened through Eq. (1). A linear extrusion
strand (see Section 2, Eq. (3)), can be expressed as:
along the strand centreline, instead, is adopted for the core wire. A
Fs = F0 + 6cos( )Fw1 purely radial contact mode is considered among the external wires
and the core, as it is usual in typical strand constructions and
Ms = M0 + 6( cos( )Mw1 + sin( )Mw3) + 6sin( )RFw1 (12a, b)
consistent with the mechanical model presented in Section 3. As a
where F0 and M0 are, respectively, the axial force and the torsional consequence, the core diameter is assumed to be greater than the
moment of the core wire. one of the external wires and the lay angle is always smaller
Eqs. (4)(12) completely dene the elasticplastic axialtor- than the maximum value max dened in (3).
sional behaviour of the strand cross section. They allow for the The 3D geometric model of the strand is then imported in
direct evaluation of Fs and Ms , if the evolution in time of the strand ANSYS for the mesh generation and the denition of the me-
axial strain and torsional curvature is known. For assigned values chanical properties of the nite element model. Twenty-node
of the axial force and torsional moment, on the other hand, the set brick element with quadratic displacement behaviour (SOLID 186
of cross sectional constitutive equations can be iteratively solved [1]) is adopted, along with the elasticplastic constitutive law al-
for s and s . To this aim, a NewtonRaphson solver is adopted in ready described in Section 3. Preliminary mesh sensitivity analyses
this work along with a classic one-step backward Euler algorithm have been carried out to identify an appropriate mesh density [6].
(see e.g. [5]) to integrate the rate-form elasticplastic constitutive A good trade-off between accuracy of results and computational
equations (8)(10) over a regular grid of concentric control points cost is found by setting the maximum size of the elements equal to
on the wire cross section. The density of the grid is controlled by 1/5 of the wire diameter in the radial direction, and to 1/40 of the
specifying in the polar coordinates of the wire ( r, ) the radial and lay length in the longitudinal direction. A medium intensity
angular discretization intervals r and . The stresses at the smoothing is also used to reduce mesh distortion [1]. Examples of
control points are then used to evaluate the wire stress resultants, the FE mesh are shown in Fig. 5 for different values of lay angle .

Fig. 5. Examples of 3D FE meshes for different values of the lay angle .


F. Foti, A. de Luca di Roseto / International Journal of Mechanical Sciences 115-116 (2016) 202214 207

Surface-to-surface contact pairs are used to model the internal restrained against rotations (xed-end case). In the latter case, the
contact among wires, accounting for both deformability of contact reacting torque is measured to quantify the coupling between
surfaces as well as possible relative sliding between wires. Friction axial and torsional behaviour.
is modelled through the classic Coulomb law. Preliminary tests, The testing procedure allows us to simulate the theoretical
however, have shown that the global response of the FE strand condition of constant axial force and torsional moment along the
model to axialtorsional loads is not affected in practice by the specimen, which can be very conveniently studied by means of the
value of the interwire friction coefcient [6]. The latter nding is in proposed analytical cross sectional model. Different sets of control
good agreement with other literature results (see e.g. [10]). Each points on the wire cross sections, with populations ranging from
contact pair is modelled with CONTA 174 and TARGE 170 elements, 25 (r = d/10, = 60) to 649 (r = d/20, = 5) points, have
while an Augmented Lagrangian algorithm is adopted to enforce been preliminary tested, showing a very fast convergence of the
contact compatibility conditions. The latter has been preferred to a integration technique adopted to evaluate the wire stress re-
Pure Penalty algorithm, in order to minimize the sensitivity of the sultants dened in Eqs. (11a, b, c, d). In this section, the results
solution to the value of the normal contact stiffness [1]. obtained with the most rened grid of integration points will be
Two rigid surface-based constraints are created at the end sec- shown, in order to present a very detailed picture of the local
tions of the strand to couple the motion of all the wire and core stress distributions predicted by the proposed mechanical
nodes to that of a single master node. The creation of these con- formulation.
straints greatly simplify the denition of the boundary conditions, Finite element simulations, instead, are performed by model-
which can be completely characterized by imposing a generalized ling a stretch of strand with a length of 500 mm, corresponding to
force (axial force or torsional moment) or displacement (axial about ve times the lay length of the external wire, to minimize
displacement or torsional rotation) to the master node. the inuence of boundary conditions (end effects) on the solution.
A mesh of about 28 000 elements and 137 400 nodes is dened
according to the criteria presented in Section 4 (see also Fig. 5).
The geometric and material parameters of the strand are taken
5. Experimental validation and numerical applications
from [13] and listed in Table 1. The results of the proposed ana-
lytical and FE models will be compared with both the experi-
Both the elasticplastic mechanical model herein proposed, as
mental results from [27] as well as the predictions of a well-known
well as the nite element (FE) approach we propose in this work
theory developed by Costello [4].
have been validated with reference to a well-documented bench-
Differently than the mechanical model proposed in this paper,
mark, widely studied in literature with both analytical as well as
Costello's one is based on the hypothesis of linearly elastic mate-
numerical techniques (e.g. [28,13,17]) and for which reliable ex-
rial behaviour. On the other hand, the geometric nonlinearities due
perimental data are available [27]. The experimental tests were
to the radial contraction of the external and core wires (Poisson
performed by means of a tensiontorsion machine, which can
effect), and the related changes in the strand internal structure, are
apply a prescribed axial elongation to a straight strand specimen,
fully accounted for. The model is geometrically nonlinear, linear-
while measuring the corresponding axial force. One end of the
ized expressions have been proposed e.g. in [29,22]. The fully non-
strand is fully clamped, whereas the other one can be free to rotate
linear Costello's model, however, is at the base of the solution
with respect to the strand centreline (free-end case) or fully
presented by Jiang et al. [13] and reported in this paper.
A comparison among theoretical and experimental results is
Table 1 shown in Fig. 6. The proposed analytical and FE models are in
Geometric and material parameters from [13].
excellent agreement with the experimental data and allow us to
d0 (mm) d (mm) (mm) E (GPa) E (GPa) y0 (GPa) capture, both from a qualitative as well as from a quantitative
point of view, the non-linear evolution of the axial load, which is
3.94 3.73 11.80 188 188 1.54 0.3 represented in Fig. 6(a) as a function of the axial strain s . Also the
coupling between the axial force and the torsional moment,

Fig. 6. Comparison among theoretical and experimental results: (a) axial force vs. axial strain and (b) axial force vs. torsional moment (xed-end test).
208 F. Foti, A. de Luca di Roseto / International Journal of Mechanical Sciences 115-116 (2016) 202214

experimentally determined through the xed-end test, is well proposed analytical formulation shows the same accuracy of
represented by the proposed models, as it can be observed from Costello's one in predicting the direct axial stiffness of the strand,
Fig. 6(b). The relation between the axial force and torsional mo- both in the xed- as well as in the free-end case. Costello's theory,
ment delivered from both the analytical and the FE model is al- on the other hand, delivers a slightly more accurate prediction of
most linear, with a small change of slope corresponding to the the coupling parameter k2. This can be explained since Costello's
complete plasticization of the strand cross section. The latter model, accounting for the radial contraction of the wire diameters
condition can be easily identied through the abrupt change of due to the Poisson effect, is based on a better estimation of the
slope of the xed-end curve in Fig. 6(a). helix radius of the external wires. The latter can inuence the
Very good agreement is also observed among the proposed parameter k2 more than k1, as it can be also inferred from in-
models and the one by Costello over the whole elastic range of the spection of the closed-form expressions given in Appendix B. A
strand response. Due to the assumption of linearly elastic material, comparison between the values of k2 obtained through the pro-
however, Costello's model obviously fails to reproduce the non- posed analytical model and Costello's one, hence, allows us to
linear evolution of the axial force, which is governed by the pro- quantify the error introduced in the strand mechanical modelling
gressive plastic deformation of wires. Table 2 shows a comparison by neglecting the Poisson effect, which is less than 2% for the case
among the elastic theoretical and experimental values of (a) the at study.
Fs Fs
direct axial stiffness k1 = s
and (b) the coupling parameter k2 = Ms
. The effects of different torsional boundary conditions on the
All values listed in Table 2 have been obtained from the initial axialtorsional behaviour of the strand can be clearly appreciated
slope of the response curves (Figs. 6(a) and (b)), with the only from the loadstrain curves depicted in Fig. 6(a). In fact, both the
exception of those referred to the proposed analytical model, initial (elastic) stiffness as well as the load-bearing capacity of the
which can be evaluated through simple closed-form expressions strand are signicantly increased by preventing the torsional ro-
(fully reported in Appendix B). Differences with respect to the tation of the end sections (xed-end case). This can be easily ex-
experimental values of parameters k1 and k2 are less than 4% for all plained through a closer analysis of the local stress state of the
theoretical models. In particular, it can be noticed that the wires.

Table 2
Fs Fs
Comparison among theoretical and experimental values of the elastic response parameters: k1 = and k2 = .
s Ms

Intentionally left blank Proposed analytical model Proposed FE model Costello's analytical model Experimental data

Fixed-end
F 13 853 13 017 13 835 13 539
k1 = s (kN)
s
Error (%) 2.32 3.86 2.19
F 1.496 1.526 1.520 1.530
k2 = s (1/mm)
Ms
Error (%) 2.25 0.26 0.65

Free-end
Fs 8895 8775 9329 9140
k1 = (kN)
s
Error (%) 2.68 3.9 2.07

Fig. 7. Fixed-end case. Results of the FE model. Load 120 kN. (a) Normal stresses. (b) Von Mises equivalent stresses.
F. Foti, A. de Luca di Roseto / International Journal of Mechanical Sciences 115-116 (2016) 202214 209

Fig. 8. Free-end case. Results of the analytical model. Load 21.5 kN. (a) Normal stresses. (b) Von Mises equivalent stresses.

Fig. 9. Free-end case. Results of the analytical model. Load 100 kN. (a) Normal stresses. (b) Von Mises equivalent stresses.

In the xed-end case, the torsional curvature of the strand is elastic region of the strand response. The normal stresses are
( )
equal to zero s = 0 and, according to the proposed analytical substantially constant over the wire cross sections, with small
model, all wires of the strand are simply stretched (see Eqs. variations due to the normal contact pressure exerted by the ex-
(4) and (5)) and subjected to a uniform distribution of normal ternal wires on the core. Contact stresses lead to an initial locali-
stresses ( ) over their cross sections. All points of a generic wire zation of yielding in the neighbourhood of internal contact sur-
cross section, hence, reach the rst yielding stress simultaneously. faces, which, however, only slightly affect the global strand be-
This leads to a sharp transition from the elastic to the post-elastic haviour (see Figs. 6(a) and (b)).
branch of the axial-strain response curve of the strand (see Fig. 6 The application of the axial load in the free-end case, on the
(a)). Within this context, the load-bearing capacity of wires is fully other hand, generates a complex and highly non-uniform stress
exploited to sustain the applied axial load. Special care, however, state over the wire cross sections. Figs. 811 show the normal ( )
and Von Mises (eq ) stress distributions obtained through the
must be paid to ensure that the end constraints are able to resist
proposed analytical (Figs. 8 and 9) and FE (Figs. 10 and 11) models
the coupling reacting moment which arises because of the pecu-
for the two different values of axial force Fs = 21.5 kN and
liar internal structure of the strand (Fig. 6(b)). The conclusions
Fs = 100 kN, which are representative, respectively, of the elastic
drawn from the analytical model are in very good agreement with and post-elastic region of the strand response. According to the
the results of the more rened FE model. The normal stresses ( ) proposed analytical model (see Eqs. (4) and (5)), the core wire is
( )
and the Von Mises equivalent stresses eq are shown, respec- subjected to the same axial strain and torsional curvature of the
tively, in Figs. 7(a) and (b) for an axial load of 120 kN, in the post- strand, while the external wires undergo a combination of axial
210 F. Foti, A. de Luca di Roseto / International Journal of Mechanical Sciences 115-116 (2016) 202214

Fig. 10. Free-end case. Results of the FE model. Load 21.5 kN. (a) Normal stresses. (b) Von Mises equivalent stresses.

Fig. 11. Free-end case. Results of the FE model. Load 100 kN. (a) Normal stresses. (b) Von Mises equivalent stresses.

elongation, torsion and bending about the binormal unit vector of of internal contact surfaces can be detected in the FE solution, as it
the SerretFrenet triad ( f3). Figs. 8(a) and (b) show the corre- can be clearly appreciated from Fig. 11(b). The analytical solution
sponding elastic distributions of the normal and Von Mises do not capture these local effects (see Fig. 9(b)), but yet it delivers
stresses, which are in good agreement with the ones evaluated both an acceptable estimate of the maximum stresses over the
through the FE model and depicted in Figs. 10(a) and (b). Since in wire cross sections and an excellent prediction of the global strand
the free-end case a portion of the external work done by the axial response curve (Fig. 6(a)).
load Fs is spent to twist and bend the wires, the initial (elastic) Additional analyses have been carried out to check the accuracy
stiffness of the strand is lower than in the xed-end case. of the proposed analytical formulation against the richer FE model
Plastic ow under the combined action of normal and tan- for different strand constructions. In all the cases, the same geo-
gential stress components, then, starts from the points where the metrical and material properties listed in Table 1 are assumed for
Von Mises equivalent stresses are highest and gradually spreads the core and external wires. Three different values of lay angle are
over the wire cross sections, thus leading to a smooth transition considered, namely, 5, 10 and 15. The highest one is close to the
from the elastic to the elasticplastic response, as it can be clearly maximum value max = 15.4, which corresponds to initial cir-
appreciated from Fig. 6(a). Analytical and FE stress distributions cumferential contact conditions and can be calculated through Eq.
compare quite well also in the elasticplastic regime, as it can be (3). Circumferential contact is not considered in the present work,
inferred from Figs. 9 and 11. Local yielding in the neighbourhood since, as already mentioned in Section 2, typical strand
F. Foti, A. de Luca di Roseto / International Journal of Mechanical Sciences 115-116 (2016) 202214 211

Fig. 12. Comparison among the results of the analytical and FE models for different values of lay angle . Axial force vs. axial strain: (a) xed-end case and (b) free-end case.

needed to prevent torsional rotations for strands with larger lay


angles. A small change of the slope, as already discussed with
reference to the previously studied benchmark problem, can be
detected both in the analytical as well as in the FE model when the
axial load reaches the value corresponding to the knee of the load
strain curve (see Fig. 12(a)). Excellent agreement is observed
among the results of the analytical and FE models, with maximum
differences on the average slope of the load-torque curve in the
order of 5%.
The proposed analytical formulation has been also compared,
in the elastic range of deformation, to a recent analytical model
developed by Argatov [2]. The latter is based on the linearly elastic
curved thin rod theory and on a rened interwire contact model,
accounting for the radial contraction of the strand due to both the
Poisson effect and the wire attening.
F
The direct axial stiffness k1 = s predicted by the two different
s
analytical models is plotted in Fig. 14(a) as a function of the lay
angle . The results are in excellent agreement, with maximum
differences in the order of 3% for the xed-end case and 4% for the
free-end case. The two models also deliver very close values of the
Fig. 13. Comparison among the results of the analytical and FE models for different F
coupling parameter k2 = Ms (xed-end case), shown in Fig. 14
values of lay angle . Axial force vs. torsional moment in the xed-end case. s
(b) for different values of lay angle. These results lead to the
constructions are characterized by interwire clearances. FE models conclusion that, for the cases considered in this work, neglecting
have been generated for each lay angle by considering a 500 mm the radial contraction of the strand, while simplifying the me-
length strand, similarly as for the benchmark test case previously chanical formulation still leads to a good estimate of the elastic
studied. stiffness parameters of the element.
The strand axial force Fs is plotted against the axial strain s in
Figs. 12(a) and (b), respectively, for the xed- and the free-end
case. Results of the analytical and FE models are in very good 6. Conclusions
agreement, for both boundary conditions, over the whole range of
lay angles herein considered. It is interesting to note that the The paper investigates the elasticplastic behaviour of metallic
variations of the lay angle slightly affect the loadstrain curve of strands under axialtorsional loads, through the combined appli-
the strand in the xed-end case. On the other hand, in the free-end cation of an innovative analytical approach and a three-dimen-
case a variation from 5 to 15 of the lay angle leads to a dramatic sional (3D) nite element (FE) model.
decrease of both the initial (elastic) stiffness and the load-bearing A new mechanical formulation is developed in this work. The
capacity of the strand. strand is regarded as a composite structural element and each
Fig. 13 shows the relation between the axial load and the re- wire is individually modelled as a curved thin rod, which can ex-
acting torque, in the xed-end case, for the different values of lay perience complex stress states due to a combination of axial
angle herein considered. The relation between axial load and elongation, torsion and bending. The evolution of plastic de-
torsional moment is almost linear with decreasing slope for in- formations in the wires is fully considered, in order to provide an
creasing values of the lay angle, i.e. larger reacting moment are effective engineering tool for the analysis of the strand response
212 F. Foti, A. de Luca di Roseto / International Journal of Mechanical Sciences 115-116 (2016) 202214

Fs
Fig. 14. Comparison among the results of different analytical formulations for different values of lay angle : (a) direct axial stiffness k1 = and (b) coupling parameter
F s
k2 = s (xed-end case).
Ms

under both service as well as ultimate loading conditions. Simple mechanical formulation herein proposed can be successfully ap-
and accurate constitutive equations are derived to relate the plied to obtain straightforward analytical estimates of both the
coupled axial load and torsional moment of the strand to the elastic stiffness parameters as well as the load-bearing capacity of
work-conjugated axial elongation and torsional curvature. the strand. The proposed approach, hence, provides some effective
The performance of the proposed analytical formulation is as- tools for engineers involved in strand design to reduce the need
sessed through comparisons both with experimental and theore-
for expensive and time consuming numerical simulations based on
tical results from the literature as well as with the numerical
non-linear 3D FE models. Moreover, the mechanical framework
predictions of a full 3D non-linear FE model. The FE model is de-
adopted in this work paves the way for the implementation of the
ned in this work starting from a parametric description of the
strand internal geometry, aiming to provide a means for a com- proposed cross sectional constitutive law in nite element for-
parative assessment of the performance of the proposed for- mulations suitable for large-scale structural analyses. This can lead
mulation over a signicant range of typical strand construction to signicant renements of the modelling strategies currently
parameters. adopted to investigate the post-elastic behaviour of cable struc-
Excellent agreement is found among the experimental results tures, which are often based on uni-axial phenomenological con-
and the proposed analytical and FE models, which allow us to stitutive laws.
describe, both from a qualitative as well as from a quantitative
point of view, the evolution of the elasticplastic response curves
of the strand. Also the experimentally observed coupling between Acknowledgements
the axial force and the torsional moment is very well described by
the proposed models. Furthermore, the analytical and FE estimates The authors are grateful to Professors Luca Martinelli and
of the elastic stiffness terms of the strand are shown to be in ex- Federico Perotti of the Department of Civil and Environmental
cellent agreement with both the experimental values and the
Engineering of Politecnico di Milano for their precious advices and
predictions of well-established linearly elastic literature models.
support.
Systematic comparisons among the analytical and FE results are
also carried out to show the ability of the proposed mechanical
approach in reproducing the local stress state of the wires, both for
free as well as for xed torsional rotations of the strand end
Appendix A. Evaluation of the wire mechanical curvatures
sections.
Additional comparisons between the analytical and FE models
are then performed to carefully assess the validity of the proposed
The mechanical curvatures of the wires, wi i = 1, 2, 3 , are ( )
mechanical formulation over a signicant range of strand internal dened, consistently with the principle of virtual work, through
geometries. Within this context, the effects on the strand beha- the expressions rst proposed in [11] and re-stated by Foti and
viour of (a) different lay angles and (b) different torsional Martinelli [8] as:
boundary conditions are highlighted. In particular, it is found that
dw
variations of the lay angle slightly affect the strand loadstrain w = + ww
dS (A.1)
curve whenever torsional rotations of the cross sections are pre-
vented. The reacting moment due to the axialtorsional coupling 3
where w is a curvature vector dened as w = i = 1 wi fi ; w is a
of the strand, however, can increase signicantly for increasing vector collecting the wire cross sectional rotations with respect to
values of the lay angle. Whenever torsional rotations are not the SerretFrenet unit vectors { fi}; S is an arc-length coordinate
constrained, instead, variations of the lay angle can dramatically
dened on the wire centreline; and w is a skew-symmetric
affect the load-bearing capacity of the strand, which is found to
decrease for increasing values of the lay angle. tensor, whose components with respect to the basis { f }
( w, ij ) i
The applications presented in the paper show that the new can be expressed as:
F. Foti, A. de Luca di Roseto / International Journal of Mechanical Sciences 115-116 (2016) 202214 213

0 0

[ w, ] = 0
ij EIw0 EI
0 0 (A.2) GJ = + 6 w (cos3( ) + (1 + )sin2( )cos( ))
1 + 0 1+
The symbols and in (A.2) denote the initial curvature and + 6sin2( )cos( )R2EA w (B.3)
torsion of the wire centreline, already dened in Section 2.
The torsional rotation of the strand can be conveniently de-
scribed, within the strand-attached reference system (SRS), by
CT = 6sin( )cos2( )REA w (B.4)
( )
means of the vector s x1 = s1e1. Then, by assuming that the cross
sections of the wires rigidly rotate with the cross section of the where EAw and EIw are, respectively, the axial and bending stiff-
strand, the vector w can be evaluated as: ness of the wire cross sections, is the Poisson coefcient of the
material, and the subscript 0 is adopted to denote quantities
w = Tws (A.3) referred to the core wire.
F
where w is the rotation tensor giving the orientation of the Expressions for the ratio k1 = s can be easily derived from (B.1)
s
SerretFrenet unit vectors { fi} with respect to the strand SRS (see ( )
for both the xed-end s = 0 as well as the free-end s 0 case: ( )
also Eq. (2)).
To obtain the wire mechanical curvatures, according to the k1 = EA (fixed end)
denition (A.1), it is necessary to derive (A.3) with respect to the CT2
k1 = EA (free end)
arc-length coordinate S. Accounting for the differential relation GJ (B.5a, b)
dS cos( ) = dx1 (see [7,8] for more details), the following expression
Fs
can be easily obtained from (A.3): Finally, in the xed-end case the ratio k2 = Ms
can be expressed
as:
w d Tw
= cos( ) Tws + ww EA
S d (A.4) k2 =
CT (B.6)
ds
where s is the strand curvature vector s = s e1, with s = dx1
.
The derivative with respect to S of the rotation tensor w gives
the variation of the orientation of the SerretFrenet unit vectors
References
along the helicoidal wire centreline. This variation is governed by
the well-known SerretFrenet formulae (see e.g. [21]), which can
[1] Ansys Inc. ANSYS mechanical user's guide. Release 15.0; November 2013.
be expressed as: [2] Argatov I. Response of a wire rope strand to axial and torsional loads:
asymptotic modeling of the effect of interwire contact deformations. Int J
dw Solids Struct 2011;48:141323.
= w w
dS (A.5) [3] Cardou A, Jolicoeur C. Mechanical models of helical strands. Appl Mech Rev
(ASME) 1997;50:114.
By recalling the properties of skew-symmetric and rotation [4] Costello GA. Theory of wire ropes.New York: Springer-Verlag; 1990.
tensors, and substituting (A.5) in (A.4), then, the following equa- [5] Criseld MA. Non linear nite element analysis of solids and structures. Vo-
lume 1: essentials. Volume 2: advanced topics. New York: Wiley and Sons;
tion can be obtained: 1997.
w [6] de Luca di Roseto A. Modellazione numerica della risposta ciclica assiale-tor-
= cos( ) Tws ww sionale e essionale di una fune metallica 1-6 [Master thesis]. Politecnico di
S (A.6) Milano, Milano, Italian; 2015.
[7] Foti F. A co-rotational beam element and a rened mechanical model for the
Finally, by substituting (A.6) in (A.1), the mechanical curvatures nonlinear dynamic analysis of cables [Doctoral dissertation]. Politecnico di
of the wire can be expressed through the following general Milano, Milano; 2013.
[8] Foti F, Martinelli L. Mechanical modeling of metallic strands subjected to
equation, rstly presented in [8]: tension, torsion and bending. Theory and nite element implementation. Int J
Solids Struct 2016;91:117.
w = cos( ) Tws (A.7) [9] Feyrer K. Wire ropes: tension, endurance, reliability.Berlin, Heidelberg:
Springer-Verlag; 2007.
The equation above leads, for the case of pure torsion of the [10] Ghoreishi SR, Tanguy M, Cartraud P, Davies P. Validity and limitations of linear
strand (i.e. s = s e1), to Eqs. (5a, b, c) introduced in Section 3. analytical models for steel wire strands under axial loading using a 3D FE
model. Int J Mech Sci 2007;49:125161.
[11] Huang NC. Theories of elastic slender curved rods. J Appl Math Phys (ZAMP)
1973;24:119.
[12] Imrak CE, Erdnmez C. On the problem of wire rope model generation with
Appendix B. Elastic stiffness coefcients
axial load. Math Comput Appl 2010;15:25968.
[13] Jiang WG, Yao MS, Walton JM. A concise nite element model for simple
Under the assumption of linearly elastic material, the cross straight wire rope strand. Int J Mech Sci 1999;41:14361.
[14] Jiang WG, Henshall JL, Walton JM. A concise nite element model for three-
sectional constitutive equations of the strand (see Eqs. (12a, b))
layered straight wire rope strand. Int J Mech Sci 2000;42:6386.
can be re-written as: [15] Jiang WG, Warby MK, Henshall JL. Statically indeterminate contacts in axially
loaded wire strand. Eur J Mech A/Solids 2008;27:6978.
Fs = EAs + CT s [16] Jonatowski JJ, Birnstiel C. Inelastic stiffened suspension space structures. J
Struct Div (ASCE) 1970;96:114366.
Ms = CT s + CT s (B.1a, b) [17] Judge R, Yang Z, Jones SW, Beattie G. Full 3D nite element modelling of spiral
strand cables. Constr Build Mater 2012;35:4529.
where EA and GJ denote, respectively, the direct axial and torsional [18] Kassimali A, Parsi-Feraidoonian H. Strength of cable trusses under combined
loads. J Struct Eng 1987;113:90724.
stiffness of the strand, while CT is the axialtorsional coupling
[19] Kmet S, Mojdis M. Postelastic analysis of cable trusses. J Struct Eng. 2015; 141
stiffness term. (10), http://dx.doi.org/10.1061/(ASCE)ST.1943-541x.0001227.
The following expressions can be obtained for the elastic [20] Knapp RH. Derivation of a new stiffness matrix for helically armoured cables
considering tension and torsion. Int J Numer Methods Eng 1979;14:51529.
stiffness coefcients [7]:
[21] Kreyszig E. Differential geometry.New York: Dover Publications; 1991.
[22] Kumar K, Botsis J. Contact stresses in multilayered strands under tension and
EA = EA w0 + 6cos3( )EA w (B.2) torsion. Trans ASME 2001;68:43240.
214 F. Foti, A. de Luca di Roseto / International Journal of Mechanical Sciences 115-116 (2016) 202214

[23] Lanteigne J. Theoretical estimation of the response of helically armoured Part I: experimental results and theoretical predictions. Int J Mech Sci
cables to tension, torsion and bending. J Appl Mech Div (ASME) 1985;52:423 1987;29:60519.
32. [28] Utting WS, Jones N. The response of wire rope strands to axial tensile loads
[24] Lemaitre J, Chaboche JL. Mechanics of solid materials. United Kingdom: Part II: comparisons of experimental results and theoretical predictions. Int J
Cambridge University Press; 1990. Mech Sci 1987;29:62136.
[25] Love AEH. A treatise on the mathematical theory of elasticity. 4th ed.. New [29] Velinsky SA, Anderson GL, Costello GA. Wire rope with complex cross section.
York: Dover Publications; 1944. J Eng Mech (ASCE) 1984;110:38091.
[26] Raoof M, Kraincanic I. Critical examination of various approaches used for [30] Yu Y, Chen Z, Liu H, Wang X. Finite element study of behavior and interface
analysing helical cables. J Strain Anal 1994;29:4355. force conditions of seven-wire strand under axial and lateral loading. Constr
[27] Utting WS, Jones N. The response of wire rope strands to axial tensile loads Build Mater 2014;66:108.