Engineering Structures 33 (2011) 1698–1709

Contents lists available at ScienceDirect

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

Numerical procedure for the analysis of damaged polyester ropes

Juan Felipe Beltrán a , Eric B. Williamson b,∗
a
Department of Civil Engineering, University of Chile, Blanco Encalada 2002, Santiago, Chile
b
Department of Civil Engineering, The University of Texas at Austin, Austin, TX 78712, USA

article info abstract

Article history: In this paper, a numerical model capable of representing the impact of damage on the overall response
Received 11 July 2010 of jacketed polyester (PET) ropes is presented. This model considers rope behavior under axisymmetric
Received in revised form loading conditions and can account for the degradation of rope properties, including the effect of broken
5 January 2011
rope components, on overall rope response. To consider the confinement effect of rope jacketing on
Accepted 8 February 2011
Available online 11 March 2011
interior rope components, the rope jacket is assumed to behave like a thin-walled tube. If a rope has
broken rope components, the rope analysis is linearized by discretizing the rope into a series of two-
Keywords:
noded axial-torsional elements. The formulation of these elements accounts for material and geometric
Damage nonlinearities. Results of numerical simulations are provided to show how changes in the L/d ratio (rope
Synthetic-fiber rope length/rope diameter) and the initial state of rope damage influence the capacity, failure strain and
Analytical model stiffness of a damaged PET rope. Experimental data are used to validate the proposed model.
Rope failure © 2011 Elsevier Ltd. All rights reserved.
Rope capacity

1. Introduction
Ropes are characterized by having a high axial strength
and stiffness in relation to their weight, combined with a low
flexural stiffness. This combination is achieved by using a large
number of components, each of which is continuous throughout
a rope’s length. When loaded axially, each component provides
tensile strength and stiffness, but when deformed in bending, the
components have a low combined bending stiffness (provided
their bending deformation is uncoupled from the axial response).
To facilitate handling, it is necessary to ensure that a rope has
some integrity as a structure, rather than being merely a set of
parallel components. This characteristic is achieved by twisting the
components together [1].
A rope can be a critical load carrying member in many engi-
neering applications, including cranes, lifts, mine hoisting, bridges,
cableways, electrical conductors, offshore mooring systems, and
so on [1–3]. Different classes of ropes, suited for different pur-
poses, have a different number and arrangement of rope compo-
nents within a rope cross-section, and rope components can be
made from a variety of different materials. Although a rope is in-
tended primarily to carry a tensile load, the construction of a rope
is such that individual rope components are subjected to bending
and torsional moments, frictional and bearing loads, as well as ten-
sion. The magnitude and distribution of the stresses resulting from
∗ Corresponding author.
E-mail address: ewilliamson@mail.utexas.edu (E.B. Williamson).
0141-0296/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2011.02.007
these loading effects determine the overall rope response, which
can be expressed in terms of the rope extension and rotation [2].
For many years, steel wire ropes have been used extensively for
load bearing members mainly due to the strength offered by steel,
coupled with the flexibility of rope construction, rope geometry,
and wire size that can be suited to a particular application. Over
the last few decades, however, the rope industry has been capable
of producing fiber ropes with very high tenacity. Considering the
many advancements made in fiber rope construction, synthetic-
fiber ropes can potentially replace steel wire ropes for certain
applications. The major difference between synthetic-fiber rope
and steel wire rope components is their strength-to-weight ratio.
Synthetic-fiber ropes can be up to 10 times lighter than a steel wire
rope for a given member strength, reducing the weight (load) of the
system and installation costs [3]. Cases where synthetic-fiber ropes
have replaced steel wire ropes are numerous and include the use
of ropes for lifting equipment and materials at construction sites,
deep water mooring systems, anti-collision and protective nets,
heavy-lift helicopters used in military and commercial operations,
and ship assist lines [3].
Each field of rope application has developed a specific body
of knowledge, based on extensive testing and field experience,
leading to empirical rules for each particular application. Unifying
these empirical rules under some general mathematical and
mechanics-based theory would allow a better understanding, and
in the long term, a better prediction of the mechanical behavior of
ropes than current methods allow. In addition, a unified modeling
approach can help reduce the need for expensive tests under a
variety of operating conditions. Thus, due to their extensive use
 J.F. Beltrán, E.B. Williamson / Engineering Structures 33 (2011) 1698–1709
Fig. 1. Cross-section of multi-level rope.
and the need to predict their behavior, several researchers have
presented analytical models that permit the calculation of rope
response based on the material properties and the geometrical
arrangement of the rope components [4,5].
In the particular case of synthetic-fiber ropes, one of the
major concerns of the rope industry is the impact of damage
on overall rope performance. Damage to ropes is a complicated
process that could depend on a variety of factors such as
strain range, abrasion, installation procedures, environmental
interaction, handling, creep, number of loading cycles, etc. ([6–9]
; among others). The process by which damage occurs can be
represented through a degradation of the properties of individual
rope components, and it can also include the complete rupture
of one or more components. Information on the residual strength
and deformation capacity of damaged polyester (PET) ropes is
important to the rope industry for developing guidelines for
estimating allowable service life and determining whether or not
damaged ropes can be retained in service.
The development of mathematical models to estimate rope re-
sponse under a variety of loading conditions that can account for
the degradation of mechanical properties has been a major focus
of research in the last few decades [9,10]. In particular, Beltran and
Williamson [11,12] have developed a numerical model to simulate
synthetic-fiber rope response under axisymmetric loading condi-
tions. This computational model, which neglects the effect of rope
self-weight, builds on previous research conducted on steel wire
ropes and synthetic-fiber ropes. Full details of the derivations are
given elsewhere ([13–15]; among others), but the main results are
summarized here for convenience. Nonlinear strain–displacement
relationships and potentially nonlinear material properties (see
Appendix for the case of PET rope components) of the components
that form a rope are considered in the model. The proposed model
has the main feature that it accounts for the degradation of rope
properties, allowing for the possibility of complete rupture of one
or more components. Degradation of rope properties is captured
by the computation of a damage index D that evolves during a
rope’s strain history and can range from 0 to 1. A value of 0 cor-
responds to the undamaged state of a rope component, and D = 1
indicates its complete rupture. For the particular case of PET ropes
under monotonically applied loads, it is assumed that the damage
index depends solely on the strain range experienced by the rope
components, and static tension tests on small-scale PET ropes (de-
fined as ropes with overall diameters between 6 and 13 mm with
corresponding maximum capacities that range between 25 kN to
62 kN) [16] validate this assumption [11,17].
The current paper provides a detailed description of the
numerical procedure implemented to account for the presence of
broken rope components on overall rope response, which extends
the capabilities of the original model. The potential confinement
effect of a rope jacket on rope components is also introduced
in this paper, and it is modeled using thin-walled tube theory.
1699
In order to compare the predicted damaged rope response with
available experimental data [18], two parameters—the number of
broken rope components and the L/d (rope length/rope diameter)
ratio—are parametrically evaluated using the proposed model to
illustrate how these two variables affect the capacity, stiffness, and
failure axial strain of a given damaged rope. In this particular study,
based on the results of prior research [17], degradation of rope
properties is represented by a power law for the damage index
evolution (see Appendix).
2. Numerical algorithm to simulate damaged rope response
2.1. Rope cross-section modeling
Full details of PET rope cross-section modeling are given
elsewhere [11,15] and only the main concepts are presented below
for completeness. In general, a synthetic-fiber rope is defined as a
structural member constructed by assembling all components in
hierarchical order. In the terminology used for this research, levels
refer to the number of structures in the hierarchy of components
that exist in a rope, and layers refer to the number of concentric
circles of components wrapped around the core. For example,
Fig. 1 shows the cross-section of a five-level, two-layer rope, along
with the parameters needed to characterize the geometry of any
individual rope component (helix radius (a), pitch distance (p),
and helix angle (θ )). The structures considered in Fig. 1 are: fibers
(j = 5), rope yarn (j = 4), strand (j = 3), sub-rope (j = 2), and rope
(j = 1). Thus, the notation adopted for the current study indicates
that level j includes all components in level j + 1, where the entire
rope is defined as level one (j = 1). Furthermore, the core is always
considered to be the first layer. In the context of Fig. 1, the six sub-
ropes around the core comprise the second layer of the rope.
Rope components, in general, are subjected to radial and
circumferential normal contact forces as shown in Fig. 2. In this
figure, the gi+1,i and gi−1,i are the line inter-layer contact forces
exerted on a rope component located in layer i by rope components
of layer i + 1 and i − 1, respectively; pnc ,i is the line inter-component
contact force exerted on a rope component located in layer i by the
other components of the same layer i; and Xi is the is the radial line
body force necessary to preserve the helical geometry of a rope
component. For computational purposes, however, two types of
cross-sectional arrangements of the components are considered:
packing geometry and wedging geometry, which represent the
extreme cases of transverse deformation for the cross-sections of
real ropes [15].
In packing geometry, it is assumed that the structures
(i.e., fibers, rope yarn, etc.) used to form a particular rope
component are initially straight, circular in cross-section, and
transversely stiff, and a twist of a specified number of turns
is imparted to the central component. Contact between rope
1700 J.F. Beltrán, E.B. Williamson / Engineering Structures 33 (2011) 1698–1709
a
c d
Fig. 2. (a) and (b) pattern of inter-layer/inter-component contact forces in an axially loaded rope element, (c) inter-layer contact forces for packing geometry, and
(d) inter-component contact forces for wedging geometry.
components in the same level is assumed to be only in the
radial direction (inter-layer contact forces) as shown in Fig. 2(c).
With wedging geometry, the components in the same level are
allowed to deform transversely and change their shape into a
wedge or truncated wedge, which is the shape that would develop
for deformable components. Radial contact is assumed to be
negligible in comparison to the circumferential contact and is
ignored for computational purposes. However, there exists an
inter-component type of contact between rope components in the
same layer for ropes with wedging geometry (Fig. 2(d)). The angle
βi is the complement of the subtended angle ψi , which is defined
as π/ni , where ni is the number of rope components in layer i.
A multi-level rope could be modeled with different arrangements
of packing and wedging geometry at different levels, generating a
mixed rope geometry.
2.2. Damaged rope analysis
As previously stated, damage in ropes can be represented
through a degradation of the properties of individual rope
components, and it can also include the complete rupture of one or
more components. The modeling of rope degradation as a function
of loading history is presented elsewhere [11,17]. Hence, the focus
of this section is on the development of a numerical algorithm to
account for the effects of broken rope components on overall rope
behavior (breaking strain, breaking stress, and rope stiffness).
Several experimental [18–23] and theoretical [12,24–28] stud-
ies have shown that if friction develops along rope components, a
broken rope component is capable of supporting its total share of
load within a relatively short distance of the rupture location. This
length is called the recovery length [25] and is defined as the dis-
tance measured from the failure end of a broken rope component
b
to the point at which this component resumes carrying its propor-
tionate share of axial load.
Previously, the authors [12] used the recovery length concept
to evaluate and quantify the effect of rope component failure on
overall small-scale PET rope response. Details of the numerical
algorithm used to compute damaged rope response, however,
were not provided in the earlier paper. Furthermore, that initial
study ignored any potential confinement effects associated with
the rope jacket, and only the normal contact forces between rope
components were assumed to give rise to the inter-layer/inter-
component interactions that permit development of admissible
recovery lengths. Observations made from recent experimental
studies on large-scale PET damaged ropes [18] of parallel sub-
rope constructions suggest, however, that rope jacket confinement
can induce normal contact forces between rope components.
In this context, large-scale refers to ropes with diameters that
vary between 32 mm and 166 mm with corresponding specified
undamaged breaking strengths (SBS) of 35 tonnes (343.2 kN) and
700 tonnes (6864.6 kN), respectively. Details of a model that
considers the effects of rope jacket confinement on the overall
response of large-scale damaged PET ropes and the numerical
implementation of the solution procedure are provided in the
subsequent sections of this paper.
2.3. Confinement effect of rope jacket
In general, a typical rope jacket has a double braid construc-
tion [29] as shown in Fig. 3. To account for the effect of the rope
jacket on the radial pressure exerted on rope components, it is pro-
posed that the rope jacket can be modeled as a thin-walled tube
for the purposes of computing overall rope response. Raoof and
 J.F. Beltrán, E.B. Williamson / Engineering Structures 33 (2011) 1698–1709
Fig. 3. Rope jacket and sub-ropes of a parallel rope construction.
Huang [26] and Gjelsvik [30] studied the problem of determin-
ing the recovery length values of a fractured wire in a parallel-
wire strand that was under external hydrostatic pressure exerted
by a prestressed wrapping or by intermittent cable bands. In the
current study, the works previously mentioned are extended and
then used to estimate the contact forces exerted by a rope jacket
on rope components for the case of synthetic-fiber ropes, where
transverse deformation of the cross-section may occur according
to the rope cross-sectional modeling techniques described previ-
ously (i.e., packing and wedging geometry).
For modeling purposes, a rope jacket can be assumed to produce
a uniform transverse compression in a rope. To illustrate the
procedure established to estimate this effect, consider first a one-
level, four-layer rope assuming inter-layer contact between the
rope components (i.e., packing geometry) (Fig. 4(a)). In the first
layer of this rope, there is one component (core), and in the
second, third, and fourth layers, there are six, twelve and eighteen
components, respectively, wound around the core (Fig. 4(a)). The
compression forces exerted by the rope jacket on rope components
of the fourth layer are transmitted radially from components of the
fourth layer to components of the first layer (core). Hence, the core
is subjected to six symmetrically distributed compressive contact
forces g2,1 per unit length (Fig. 4(b)).
It is assumed that the rope jacket develops longitudinal and
circumferential stresses according to the assumptions of thin-
walled tube behavior (i.e., a thin-walled cylindrical pressure
vessel). If H is the circumferential line force in the rope jacket
acting at a given cross-section (Fig. 4(c)), the nominal compressive
radial stress q can be computed, assuming the rope cross-section
a b
d
Fig. 4. (a) Rope cross-section; (b) Normal forces acting on the core; (c) and (d) circumferential force and longitudinal stress acting in the rope jacket with thin-walled tube
behavior, respectively; (e) compressive stress acting on the core.
1701
is solid, by
2H
q= (1)
d
where d is the rope diameter, and the stress q is related to the
longitudinal stress in the rope jacket σL by the following expression
(Fig. 4(d)):
4σL t
q= (2)
d
where t is the thickness of the rope jacket.
The stress q would be the radial stress if the rope’s cross-section
were solid and without voids between the rope components. If
there are voids between rope components (packing geometry),
Gjelsvik [30] suggested a procedure to compute the compressive
contact forces g2,1 exerted on the core based on an equivalent
cross-section that makes the rope solid, (i.e., without the voids
between rope components (Fig. 4(e))). The equivalent cross-
section is obtained by intersecting the tangent lines at every
contact point between the core and the components wrapped
around it. For the rope cross-section shown in Fig. 4(a), the
equivalent cross-section is a hexagon that circumscribes the core
cross-section (Fig. 4(b)). Thus, considering equilibrium in the
vertical direction (Fig. 4(e)), the compressive contact forces g2,1 are
given by
 
2 dc
g2,1 = H (3)
3 cos γ1
sin γi
∑
d
i=1
where dc is the diameter of the core and the angle γi is given in
Fig. 4(e). The procedure just described can be used to obtain the
value of the compressive contact forces g2,1 for a different number
of rope components wrapped around the core.
LeClair [31] presents a study in which the contact geometry
between helical layers is established and in which equilibrium is
enforced in the radial direction at contact points. In this model,
the radial force is transmitted from the outer layers to the inner
layers, accounting for the distance between points of contact along
the length of a helical rope component. This distance between
contact points depends upon both the helix angle and the radius
of components in the layers under consideration. Thus, for i ≥
c
e
1702 J.F. Beltrán, E.B. Williamson / Engineering Structures 33 (2011) 1698–1709

2, expressions for the contact line forces (gi−1,i (see Fig. 2(c)))
transmitted from the outer layers to the inner layers can be
estimated as
 
j−1
nj pi cos αi cos αk
−ml ∏
gi−1,i = Xi + Xj (4)
n p cos αj
j=i+1 i j k=i+1
cos αk
where ml is the number of layers in the rope and nl , pl and Xl are
the number of rope components, the pitch distance (which can be
used to determine the helix angle), and the radial line body force
of a rope component in layer l, respectively. The angles αl and αl Fig. 5. Two-noded axial-torsional element.
are defined as tan−1 (pl /2π (al − rl )) and tan−1 (pl /2π (al + rl )),
respectively, where al is the helix radius and rl is the radius of a
a
rope component in layer l [32].
Eq. (4) can be used to evaluate the contact line force g1,2 due
solely to the compression forces, grj,ml , exerted by the rope jacket
(rj) on the rope components of the outermost layer (ml) of a rope.
Setting the value of Xml equal to grj,ml and eliminating the values
of Xi for i = 2, . . . , ml − 1, an expression for g1,2 is obtained.
Rope components that belong to layer two are wound around a b
central straight core; thus, the inter-layer contact between layers
one and two describes a helical path. As such, equilibrium in the
radial direction requires that the contact line forces g1,2 and g2,1 are
related by g1,2 = g2,1 cos(θ2 ), where θ2 is the helix angle associated
with the rope components in the second layer. Therefore, the
contact line force g2,1 can be estimated in terms of the compressive
contact force grj,ml as follows:
ml−1 
∏ cos αk Fig. 6. Damaged rope discretizations: (a) failure located at one end of the rope and
(nml p2 cos α2 ) cos αk (b) failure located around the midspan of the rope.
k=3
g2,1 = grj,ml . (5)
cos θ2 n2 pml cos αml length values for a general multi-level rope with packing, wedging,
Hence, or mixed geometry, considering a general nonlinear slip-stick
friction model, are not presented in this paper due to space
cos θ2 n2 pml cos αml limitations; readers are encouraged, however, to consult the work
grj,ml =  g2,1 . (6)
by Beltran and Williamson [12] for a detailed description of these
ml−1
∏ cos αk
(nml p2 cos α2 ) cos αk equations. To simplify the analysis, the gradual increase in load
k=3
carried by broken rope components is ignored. Thus, only the full
For the case of a rope cross-section modeled using wedging geom- value of axial load at distances outside admissible recovery length
etry, by assumption, there are no voids between rope components values are taken into consideration. This assumption is based
and there is no radial contact between rope components of con- on parametric studies in which the gradual (nonlinear) increase
tiguous layers. The circumferential line force H in the rope jacket in tension of a broken rope component for a particular rope
is equilibrated by the rope components located in the outermost construction was computed, considering different positions within
layer of the rope. Thus, using the same notation as before, the com- the rope cross-section and different failure axial strain values of
pressive contact force grj,ml exerted by the rope jacket on every rope the broken rope components. These studies, which assume that
component of the outermost layer (ml) can be estimated as frictional forces are linearly dependent on the normal forces (i.e., a
2H π Coloumb friction model), show that most of the tension load is
grj,ml = (7) accumulated toward the end of the recovery length interval [0, rl]
nml (nonlinear profile) or the recovery length values are relatively
where nml is the number of components in the outermost layer short [33].
of the rope being analyzed. Once the compressive contact force The recovery length of an individual broken rope component
grj,ml exerted by the rope jacket on every rope component of the is computed for a prescribed deformation level of the entire
outermost layer (ml) is obtained, the radial force transmitted from rope (ε, ζ ), where ε is the rope axial strain and ζ is the rate of
the outer layers to the inner layers and/or the contact force exerted twist. Associated with the rope axial strain ε is the rope axial
on components that belong to lower levels can be estimated by displacement u, and associated with the rate of twist of the
considering equilibrium in the radial direction as explained by rope is the rope axial rotation φ . If the recovery length value
Beltran and Williamson [12]. is admissible (i.e., within the physical length constraints of the
rope being analyzed), the problem is linearized by discretizing
2.4. Numerical model of damaged rope response the rope along its length into axial-torsional elements (Si ), with
element lengths LSi , based on the value of the recovery lengths
The proposed numerical model developed to estimate the effect computed. Each element includes two degrees of freedom at each
of broken rope components on overall rope response considers, as node: axial displacement (ui ) and axial rotation (φi ). Associated
an initial step, whether or not broken rope components can take up with these degrees of freedom are the corresponding axial force
their appropriate share of axial load by considering the equilibrium Ti and torsional moment Mi (Fig. 5).
equations that govern rope behavior and by computing their In Fig. 6, two possible rope discretizations are presented along
recovery lengths. The equations that allow computing recovery with the corresponding axial-torsional element cross-sections,
 J.F. Beltrán, E.B. Williamson / Engineering Structures 33 (2011) 1698–1709
where the broken rope components have been identified with an
‘‘X’’. Also shown in the figure are the incremental changes in the
degrees of freedom and nodal forces for each lengthwise element.
The two possible discretizations are based upon the location of the
failed rope component(s): (a) the failure region is located near one
end of the rope (i.e., at a splice location), or (b) the failure region
is located near the rope midspan. In both possible discretizations,
the length LS2 represents the damage length; for case (a) its value
is equal to rl and for case (b) its value is equal to 2 · rl. The proposed
model captures the effect of having a weakened cross-section
acting over a localized region characterized by the recovery length.
Based on equilibrium considerations, the response of an individual
axial-torsional element can be described in terms of only two
independent degrees of freedom (referred to as the ‘‘essential set’’).
Hence, based on the degrees of freedom defined for the element
shown in Fig. 5, only the axial deformation (u3 − u1 ) and the axial
rotational deformation (φ4 − φ2 ) are needed. The element forces
associated with the essential set are taken as T3 and M4 .
The tangent stiffness matrix of each axial-torsional element Si
is computed, and then each tangent stiffness matrix is assembled
to obtain the tangent stiffness matrix of the entire rope. For each
axial-torsional element, a 2 × 2 stiffness matrix is computed having
the following form:
[ ]
kλλ kλη
[ksr ]i =
kηλ kηη
i
where the subscripts λ and η are associated with the axial
strain and the rate of twist, respectively, of the cross-section
of element Si . No closed-form expressions exist to compute the
stiffness coefficients due to the nonlinear geometry and nonlinear
constitutive response of the rope components ([11] and Appendix).
Accordingly, a numerical procedure is needed to compute these
terms. Thus, if there exists a function Gi = (G1 , G2 )i such that
(Ti , Mi ) = Gi (λ, η)i , the stiffness coefficients can be obtained
by computing the gradient of the function Gi . Accordingly, the
stiffness matrix (Eq. (8)) of each axial-torsional element has the
following form:
 ∂G ∂ G1 
1
[ksr ]i =  ∂∂λ ∂η 
∂ G2  .

G2
∂λ ∂η
i
As stated before, the function Gi does not admit closed-form
expressions for any element Si . Thus, the partial derivates of
Eq. (9) are evaluated using a numerical scheme by inducing small
perturbations (δλ, δη)i to a prescribed level of deformation (λ, η)i
for each axial-torsional element Si . These induced perturbations
are applied separately (i.e., small perturbations are assumed for
one variable while holding the other perturbation equal to zero),
generating two different perturbed deformations. Because the
computational model considers small increments in axial strain
and axial rotation per unit length (δλ and δη, respectively),
the standard engineering strain definition is used for each
element Si . Using the damage-dependant constitutive equations
corresponding to the current level of rope deformation (Appendix),
the increments in axial force and twisting moment on the
axial-torsional elements due to axial deformations and twisting
deformations are then computed. These computed forces and
moments, along with the forces and moments associated with the
unperturbed deformation level (ε, ζ )i , are used to compute the
stiffness coefficients. Accordingly, the stiffness coefficients of the
matrix [ksr ]i can be evaluated, for any element Si , as follows:
∂ G1 T (λ + δλ, η) − T (λ − δλ, η)
[ ]
kλλ = =
∂λ 2δλ
1703
∂ G2 M (λ + δλ, η) − M (λ − δλ, η)
[ ]
kηλ = = (10b)
∂λ 2δλ
∂ G1 T (λ, η + δη) − T (λ, η − δη)
[ ]
kλη = = (10c)
∂η 2δη
∂ G2 M (λ, η + δη) − M (λ, η − δη)
[ ]
kηη = = (10d)
∂η 2δη
where the centered differentiation formula is used to evaluate the
stiffness coefficients, T and M are the axial force and torsional
moment, respectively, and the values of the perturbations δλ
and δη are found using an iterative process where variations are
considered until two consecutive values of the stiffness coefficients
are sufficiently close based upon a prescribed tolerance. The values
of T and M are computed using the formulation described in [11].
The 4 × 4 tangent stiffness matrix [Ksr ]i of each axial-torsional
element Si , associated with the degrees of freedom u1 , φ2 , u3 and
φ4 (referred to as the ‘‘complete set’’) described in Fig. 5, can be
obtained through consideration of equilibrium. Hence, with the
2 × 2 element stiffness matrix [ksr ]i computed, [Ksr ]i can be
calculated as follows:
[ ]
[ksr ]i −[ksr ]i
[Ksr ]i = . (11)
−[ksr ]i [ksr ]i
A displacement control analysis is carried out to obtain overall
(8)
damaged rope response. Increments in rope axial strain, dεj , and
rope axial rotation per unit length, dζj , are specified for the entire
rope in the jth incremental step of the analysis. The aim of the
analysis is then to obtain the values of the increments of axial
displacements (dui )j and axial rotations (dφi )j of the individual
axial-torsional rope elements, as well as the increments in the
external loads (axial forces and torsional moments), required to
produce the specified increment in the deformation level (dεj , dζj )
of the rope. These quantities are obtained by solving the following
linear system of equations:
[Kr ]j−1 {du}j = {dP }j (12)
where [Kr ]j−1 is the rope tangent stiffness matrix that is computed
by considering the rope configuration of the (j − 1)th incremental
step, which is obtained by assembling the tangent stiffness
(9)
matrices of each axial-torsional element computed from Eq. (11).
The value of [Kr ]j−1 is used to compute the rope configuration
at the jth incremental step because at the end of the (j −
1)th incremental step the rope has reached an equilibrium
configuration so that all the stiffness coefficients are known; {du}j
is the increment in axial displacements and axial rotations, and
{dP }j is the increment in axial forces and torsional moments
in the jth incremental step. Eq. (12) represents the linearized
solution of damaged rope behavior for each increment in the
deformation level of the rope being analyzed. Because of the
nonlinearity of the system properties, unbalanced forces are
generated due to linearization; hence, an incremental-iterative
numerical procedure is used to compute the overall damaged rope
response. Therefore, Eq. (12) can be recast in the following form:
[Kr ]jk−1 {du}kj = {dQ }kj + {R }kj −1 (13)
where the superscript k represents the iterative step, [Kr ]kj −1
represents the tangent stiffness at the end of the (k − 1)th iteration,
{dQ }kj is the increment in the external load vector of the kth
iteration, and {R }kj −1 is the unbalanced load vector (axial forces
and torsional moments) that represents the imbalance between
the existing internal and external loads at the end of the (k − 1)th
iteration.
To illustrate the solution and the form of the components in
(10a)
Eq. (13), consider a rope in which two of its seven components have
1704 J.F. Beltrán, E.B. Williamson / Engineering Structures 33 (2011) 1698–1709
failed around the rope midspan and their recovery length values
are admissible. This damaged rope is discretized into three axial-
torsional elements as shown in Fig. 6(b). For this case, the system
of equations presented in Eq. (13) has the following form
 k−1  du1 k dQ1 k R1 k−1
k11 ··· k18
dφ 2 
dQ2  R2 
 .. .. ..  
 . . .   ..  =  ..  +  ..  . (14)
     
. . .  .
k81 ··· k88 j
dφ 8 j dQ8 j R8 j
For this example, the unknown increments in displacements,
rotations, axial loads, and torsional moments for each incremental
step of the analysis of Eq. (13) are du3 , dφ4 , du5 , dφ6 and
dQ1 , dQ2 , dQ7 , dQ8 , respectively. The quantities du1 = −du7 and
dφ2 = −dφ8 are known because, as stated before, a displacement-
based analysis is performed, and dQ3 , dQ4 , dQ5 , and dQ6 vanish due
to equilibrium considerations. In this particular application, the
tangent stiffness [Kr ]jk−1 is kept constant during iterations, and it is
updated at the end of each increment. For the first incremental step
(j = 1) of the analysis, the tangent stiffness [Kr ]j=1 is computed for
a small level of deformation (λ, η)i , on the order of 1E-4 for each
element Si .
At the beginning of each incremental step (i.e., k = 1), the
system is in equilibrium. As such, the unbalanced load vector {R }0j
vanishes and, as stated before, the increments in displacement
and rotation of the entire rope are prescribed. At the discretized
level of the axial-torsional elements, the unknown increments
in displacements and rotations {du3 , dφ4 , du5 , dφ6 }1j are first
computed using the third through sixth equations of the system
of equations represented by Eq. (14) using a Gaussian elimination
scheme. Having computed these quantities, the increments in
external axial loads and torsional moments {dQ1 , dQ2 , dQ7 , dQ8 }1j
are computed based on the first, second, seventh and eighth
equations of the system of equations given by Eq. (14). The
total displacements, rotations, external axial loads, and torsional
moments can be updated as {U }1j = {U }j−1 + {du}1j and {Q }1j =
{Q }j−1 + {dQ }1j where {U }1j and {Q }1j are the vectors of the current
axial displacements and axial rotations and external axial loads and
torsional moments, respectively, of the rope; {U }j−1 and {Q }j−1
are the vectors of total axial displacements and axial rotations and
total external axial loads and torsional moments, respectively, at
the end of the (j − 1)th increment.
Axial strain {λ}1j and rate of twist {η}1j for each element Si are
computed using the standard engineering strain definition. Using
the damage-dependant constitutive equations corresponding to
the current level of rope deformation (Appendix), the internal
axial force {T }1j and twisting moment {M }1j on each axial-torsional
element due to axial and twisting strains are computed. Then, the
unbalanced load vector {R }1j is computed as the difference between
the internal and the external loads. If the norm of the unbalanced
load vector {R }1j (i.e.,‖Rj1 ‖) is greater than a prescribed tolerance,
the rope is not in equilibrium, and iterations on the variables
du3 , dφ4 , du5 , dφ6 , dQ1 , dQ2 , dQ7 , and dQ8 are needed. For k ≥ 2,
the following system of equations is solved:
 0  δ(du1 ) k R1 k−1 δ(dQ1 )k
k11 ··· k18
 .. δ(dφ2 ) R2  δ(dQ2 )
.. ..   .  = .  + . 
 . . .  .  .  .  (15)
. . .

k81 ··· k88 j
δ(dφ8 ) j R8 j δ(dQ8 ) j
where the operator δ() is associated with the variations of
the increments in axial displacements, axial rotations, external
axial loads and torsional moments. As stated previously, the
values of {du1 , dφ2 , du7 , dφ8 }j are prescribed; thus, the values of
{δ(du1 ), δ(dφ2 ), δ(du7 ), δ(dφ8 )}kj vanish in Eq. (15) for k ≥ 2.
In addition, the values of {δ(dQ3 ), δ(dQ4 ), δ(dQ5 ), δ(dQ6 )}kj also
vanish in Eq. (15) for k ≥ 2 due to equilibrium considerations.
Hence, the variation in the increments in axial displacements and
axial rotations, {δ(du3 ), δ(dφ4 ), δ(du5 ), δ(dφ6 )}kj , are obtained by
solving the following system of equations:
δ(du3 ) k
 k−1
 0  
k33 ··· k36 R3
 .. .. ..  δ(dφ4 ) = R4  .
. .   δ(du5 )  R  (16)
5
k63 ··· k66 j δ(dφ6 ) R6 j
j
The variation in the increments of the external axial loads
and the torsional moments {δ(dQ1 ), δ(dQ2 ), δ(dQ7 ), δ(dQ8 )}kj are
computed based on the first, second, seventh and eighth equations
of the system of equations represented by Eq. (15). In general, the
total displacements, rotations, external axial loads, and torsional
moments in the jth incremental step of the analysis can be updated
as follows:
mi
−
{U }kj = {U }j−1 + {du}1j + {δ(du)}kj (17a)
k=2
mi
−
{Q }kj = {Q }j−1 + {dQ }1j + {δ(dQ )}kj (17b)
k=2
where {δ(du)}kj is the vector of the variation of the increments
in axial displacements and axial rotations for the kth iteration,
{δ(dQ )}kj is the vector of the variation of the increments in external
axial loads and torsional moments for the kth iteration, and mi is
the total number of iterations performed during the jth analysis
step.
Once the total axial displacements and axial rotations {U }kj
are computed, axial strain {λ}kj and rate of twist {η}kj for
each axial-torsional element Si are computed, as before, using
the standard engineering strain definition. Likewise, using the
damage-dependant constitutive equations corresponding to the
current level of rope deformation (Appendix), the internal axial
force {T }kj and twisting moment {M }kj on each axial-torsional
element due to axial deformations and twisting deformations are
computed. Then, the unbalanced load vector {R }kj is computed in
a similar manner as explained in the case of {R }1j . If the norm of
{R }kj is less than a prescribed tolerance, the iteration process is
stopped and a new analysis is performed for the (j + 1)th increment
in axial deformation (dε(j+1) ) and axial rotation (dζ(j+1) ) of the
rope. Otherwise, new iterations are performed until equilibrium
is achieved.
3. Numerical implementation
In this section, two example ropes are analyzed using the
methodology described in the previous sections. Based on the
model formulation, some of the variables that may affect damaged
rope response are the number of broken rope components and
their distribution around a rope cross-section, the L/d (rope
length/rope diameter) ratio, the location of a damaged cross-
section relative to the rope length, the potential confinement
exerted by a rope jacket, and the rope cross-sectional modeling
technique (i.e., packing or wedging geometry) [33]. Due to space
limitations and availability of experimental data [18], only the L/d
ratio and the number and distribution of broken rope components
around a rope cross-section are evaluated in the current paper
to determine their influence on the capacity, stiffness, and failure
strain of a given damaged PET rope. The L/d ratios selected in
the testing program reported by Ward, et al. [18] vary between
40 and 1000 and consider an initial damage state in which
10% of the cross-section is cut to simulate what may happen
 J.F. Beltrán, E.B. Williamson / Engineering Structures 33 (2011) 1698–1709
during installation and handling. While the data reported by Ward
et al. [18] serve as a basis for the analysis results presented
below, the primary intent of including this information is to
illustrate the implementation of the proposed numerical modeling
technique and to demonstrate how the model can be used to
evaluate damaged rope performance. As such, the numerical
results consider a broader variation in the L/d ratio (from 40 to
3000) and initial damage state (from 3% to 20%) than was reported
in the experimental testing program.
The numerical simulations and comparisons with experimental
data presented in this study consider two different types of
rope construction, with corresponding damaged PET rope cross-
sections shown in Fig. 7. These ropes were selected to correspond
to two of the specimens evaluated in a recent experimental testing
program [18] on large-scale, polyester-jacketed ropes (DPJR) with
a specified breaking strength (SBS) of 35 Tonnes (343 kN). Both
ropes have a parallel sub-rope construction, which means that the
sub-ropes are not twisted together in a helical fashion. Each sub-
rope itself, however, is built by twisting together a certain number
of strands. For analysis purposes, and to simulate the testing
program previously mentioned, damage is inflicted to each rope at
the midspan location and at the strand level prior to loading. This
damage is simulated by cutting (i.e., specifying a damage index D =
1 (see Appendix)) a certain number of strands to reduce the cross-
sectional area by a prescribed percentage. All damage is inflicted to
adjacent sub-ropes near the exterior of the rope to simulate surface
damage. For modeling purposes, the rope is analyzed as a two-
level rope: the first level of the rope (sub-rope) is treated as having
a packing geometry, and the second level of the rope (strand)
is modeled using wedging geometry if no initial damage exists
and packing geometry otherwise. This assumption is based on
results from simulations of small-scale PET damaged rope behavior
presented by Beltran and Williamson [17] and due to the fact
that if damage is present at the strand level, undamaged strands
can migrate to a new a position because of the space available
from voids between undamaged strands (i.e., the packing geometry
assumption). For each rope analyzed, the complete hierarchical
structure is given. Thus, the response of each strand is based on
the constitutive properties of the rope yarns that comprise the
strand. This information is available from test results that give
tensile load as a function of the axial strain, where the breaking
axial strain is 13% and the breaking axial load is specified for each
rope construction [34].
According to work by Beltran and Williamson [12], which
ignores the presence of a rope jacket, damaged components
within ropes that have a parallel sub-rope construction do not
develop contact forces as a result of rope axial deformations;
hence, no frictional forces exist between components. Under such
conditions, broken rope components do not develop admissible
recovery length values, and damaged rope response is computed
ignoring the contribution of broken rope components. As stated
previously (Section 2.2), however, Ward et al. [18] suggest that
rope jacket confinement can induce normal contact forces between
rope components. Thus, broken rope components can eventually
develop admissible recovery length values and start carrying load,
which indicates that they can contribute to overall rope response.
To study this potential non-uniform strain distribution along the
length of a damaged rope, the confinement exerted on a rope by
the jacket is considered using the thin-walled tube approximation
described in Section 2.3. As such, a damaged rope analysis is carried
out by discretizing the length of the rope into axial-torsional
elements, where the lengths and the number of rope components
for each axial-torsional element are based on the value of the
recovery lengths computed.
For each rope analyzed, it is assumed that the rope jacket
has an axial capacity that is 3% of the initially undamaged rope
1705
a b
Fig. 7. Damaged (10%) rope cross-section (a) DPJR1 and (b) DPJR2.
capacity, is comprised of the same PET material that is used to
form the components within the rope, and has a thickness that
is approximately equal to 2% of the rope diameter. It is also
assumed that the rope jacket is completely bound to the rope
(i.e., there is no slip between the rope components and the jacket).
Ayres [35] established that, for undamaged ropes, the jacket can
account for up to 8% of the theoretical rope strength. Because the
actual value is not known and has not been measured, a small
value, relative to the overall rope capacity, is assumed for the
rope jacket capacity, which is an assumption that is consistent
with the research reported by Ayres [35]. Furthermore, results
from parametric studies using different values for the assumed
axial capacity of the jacket show small differences in the overall
computed response of a rope [33]. Only for those cases in which
the jacket strength is ignored or is selected to be an unrealistically
large value do the results deviate from those presented below.
Recovery length values are computed assuming a Coulomb
friction model with a friction coefficient equal to 0.1 ([36,37]) and
a small level of rope deformation (10% of the specified failure
axial strain). Based on these selected parameters, the recovery
length of the broken rope components that belong to the rope-
cross section depicted in Fig. 7(a) (DPJR1) is equal to 820 mm,
and the recovery length of the broken rope components that
belong to the rope-cross section depicted in Fig. 7(b) (DPJR2) is
equal to 615 mm. These results indicate that damage propagates
over a longer length in DPJR1 than in DPJR2. This behavior is
due to the fact that the strands of DPJR1 are stiffer than the
strands of DPJR2 (Fig. 7), which induces a higher axial load level
in the strands of DPJR1 relative to DPJR2 for the same deformation
level (10%). This effect is not compensated by the fact that the
pressure exerted by the rope jacket of DPJR1 is distributed over
fewer components than DPJR2 (i.e., the confining pressure per
component for DPJR1 is greater than DPJR2) and the rope jacket
of DPJR1 is thicker than the rope jacket of DPJR2, thereby inducing
higher pressure on its rope components. Despite these mitigating
factors, the recovery length in DPJR1 is longer than the recovery
length in DPJR2. Consequently, for the case of DPJR1 with an L/d
ratio equal to 40, the recovery length on each side of the centrally
located failure location is greater than the total rope length, making
the recovery length value inadmissible. In this case, damaged
rope response is computed by using a net area effect model. This
model ignores the contribution of broken rope components on
overall rope response after they fail. Accordingly, the stiffness
and strength loss of a rope are proportional to the area that the
broken rope components represent with respect to the total rope
cross-sectional area. Furthermore, with the net area effect model,
the rope breaking strain remains unchanged with respect to the
initially undamaged rope failure strain [11,16].
For damaged ropes with an L/d ratio equal to 290 and 1000,
damage lengths of the ropes are approximately 15% and 4% of the
1706 J.F. Beltrán, E.B. Williamson / Engineering Structures 33 (2011) 1698–1709

corresponding overall rope length, respectively, which indicates a

that these recovery lengths are admissible. For the case of DPJR
2 and the selected values of the L/d ratio (40, 290, and 1000), all
recovery length values are admissible and correspond to damage
lengths of approximately 96%, 13%, and 3.5%, respectively, of the
corresponding overall rope length.
Fig. 8(a) shows the predicted response of DPJR1 as a function
of the L/d ratio for an assumed level of initial damage equal
to 10% of the total cross-sectional area. In this study, consistent
with the experimental data [18], the L/d ratio changes by varying
the rope length L and holding the rope diameter d fixed. For
reference, the predicted response of the initially undamaged
rope is also provided in the figure. All analyses account for the
degradation of rope properties due to axial deformation according b
to the damage-dependent constitutive model presented in the
Appendix. Numerical simulations show that as the value of the L/d
ratio increases, the failure axial strain decreases, damaged rope
response approaches the undamaged case, and the rope capacity
values appear to be unchanged. These results suggest that as rope
length increases, the impact of initial damage on overall rope
response becomes less important. In Fig. 8(b), the dependence
of the damaged rope response on the initial damage state for
an assumed value of L/d equal to 80 is evaluated. This value
of L/d was selected to clearly distinguish the response curves
of DPJR1 considering different initial damage states. For greater
L/d values, the aforementioned response curves converge to one
Fig. 8. Damaged DPJR1 response (a) for different values of L/d and 10% of initial
another as explained in the subsequent discussion. As the initial damage and (b) for different values of initial damage and a value of L/d = 80.
damage state increases, the damaged rope capacity decreases and
both the failure axial strain and the rope stiffness decrease. The
response trends identified in Fig. 8 can be extended to evaluate a
ropes with other combinations of L/d ratio and initial damage.
Thus, considering that the damaged rope stiffness approaches the
undamaged case as the L/d ratio increases (Fig. 8(a)) and that the
proposed model suggests (Fig. 8(a) and further discussion below)
that the damaged rope capacity does not depend strongly on the
L/d ratio, the analysis results presented in Fig. 8(b) would approach
those of the initially undamaged rope response curve if the value
of the L/d ratio were increased. However, rope failure strain would
be different for each curve considering its initial damage state—as
the initial damage state increases, rope failure strain gets smaller.
These factors are considered further in the discussion below.
In Fig. 9, the predicted rope capacities and failure strains of
DPJR1 for varying levels of initial damage are plotted as a function b
of the L/d ratio. The predicted rope capacities (Fig. 9(a)) do not
show a strong dependence on the L/d ratio for all the initial damage
states considered, which means that the capacity reduction in
damaged ropes depends primarily on the initial loss of cross-
sectional area. This behavior is a direct result of the assumptions
used to formulate the proposed analysis model. Based on these
assumptions and the fact that loss of symmetry due to rope
component failures is neglected, the damaged rope capacity is
controlled by the capacity of the weakened cross-section of the
rope that acts over a localized region represented by an individual
axial-torsional element as explained in Section 2.4. The predicted
failure axial strains for a given level of initial damage (Fig. 9(b)),
however, do show a dependence on the L/d ratio, with a decreasing Fig. 9. Variation of (a) failure axial load and (b) failure axial strain with respect to
rate of change as the L/d ratio increases. Eventually, for all cases L/d ratio for different initial rope damage states of rope DPJR1.
studied, the failure axial strains reach a minimum value for large
values of the L/d ratio, and this minimum value depends upon the axial-torsional elements if the value of its recovery length is
initial level of damage. In addition, the rate of convergence to this admissible, where the central element (Fig. 6(b)) is the localized
minimum value is faster for smaller values of initial damage. region where strain and damage concentrate along the rope
In evaluating the computed results associated with DPJR1 length [12]. Considering that the value of the recovery length is
(Figs. 8 and 9), it is important to note that the recovery length constant for a given distribution of cross-sectional damage, as
is independent of the rope length and solely depends on the the L/d ratio gets larger, the percentage of the damaged length
rope cross-sectional model [12]. Based on the formulation of the with respect to the overall rope length gets smaller, and the sum
proposed model, rope DPJR1 is discretized into three two-noded of the lengths of the axial-torsional elements that consider the
 J.F. Beltrán, E.B. Williamson / Engineering Structures 33 (2011) 1698–1709
a damage state corresponding to 10% of the cross-sectional area of
b of DPJR2, the failure axial load also increases (about 5%) as the
Fig. 10. Comparisons between experimental and predicted values of (a) failure
axial load and (b) failure axial strain considering a 10% of initial damage state.
contribution of all rope components in the overall rope response
(LS1 and LS3 in Fig. 6(b)) approaches the actual rope length.
Hence, the damaged rope stiffness approaches the stiffness of
the initially undamaged rope and, as previously stated, the rope
failure strain converges to a minimum value. Accordingly, for ropes
with a given damage state in which the damage length is a small
percentage of the rope length, the reduction of the failure axial
strain with respect to the undamaged case can be considered to be
its minimum value. Likewise, it should also be noted that overall
damaged rope response shows a similar dependency on the axial
capacity of the rope jacket, which determines the recovery length
values. Thus, if jacket axial capacity is increased for a given number
of broken rope components, the recovery length becomes shorter,
indicating that the percentage of damaged length compared to the
overall rope length becomes smaller; hence, the same performance
trends as identified above are observed [33]. In addition, as the
initial loss of rope cross-sectional area increases for a given L/d
ratio, the predicted failure axial strain decreases. This result is
due to the fact that as the initial loss of rope cross-sectional area
gets larger, the localized damage region of the rope defined by the
recovery length also increases [12]. For example, based on Fig. 9(b),
the analyzed rope (DPJR1) with initial damage states of 3%, 10%,
and 20% has a reduction in its failure axial strain by approximately
7%, 15%, and 24%, respectively, relative to the undamaged case
(rope failure strain equal to 0.13) for a value of the L/d ratio equal
to 3000.
In Fig. 10, comparisons between experimental data and
predicted values of the failure axial load and failure axial strain
are presented for DPJR1 and DPJR2. Both of these ropes (and all
the ropes tested in the study reported by Ward [18]) were made
of polyester fibers obtained from a single manufacturer. Thus,
for computational purposes, the same constitutive function that
describes the tensile load as a function of the axial strain of the
polyester textile yarns was used for all the ropes analyzed in this
paper (see Appendix). The results shown in Fig. 10 are for an initial
1707
each rope, which was the initial damage state value considered in
the testing program [18]. Not only are the values of the failure load
and failure strain compared, their dependence on the L/d ratio is
also provided. For each rope tested, only the values of the rope
breaking strain and rope breaking load were reported [18]. No
information was given about the gage length used to compute the
values of rope breaking strain obtained from the rope elongation
measurements.
Failure axial loads for both ropes analyzed (Fig. 10(a)) are
accurately predicted by the proposed numerical model (within a
range of −5% to +10% of the experimental values). Experimental
data indicate that for DPJR1, the failure axial load increases (about
7%) as the L/d ratio increases from 40 to 560, but it remains nearly
constant for values of L/dequal to 560 and 1000. For the case
L/d ratio increases from 40 to 290, but then it decreases (about
18%) as the value of L/dincreases from 290 to 1000. Because the
experimental results between DPJR1 and DPJR2 are not consistent
and do not show a clear trend in how their capacity depends
upon the L/d ratio, the computed results, which show limited
dependence on the L/d ratio, are considered to be reasonable.
As shown in Fig. 10(b), predicted values for the rope failure
axial strain overestimate the experimental values for the two cases
analyzed. Specifically, experimental values for the failure axial
strain are overestimated within a range of 30% to 40% for DPJR1
and 40% to 60% for DPJR2 relative to the experimental values over
the entire range of L/d ratios considered. In Fig. 10(b), as expected,
DPJR2 reaches its minimum failure axial strain value for a smaller
value of rope length L than DPJR1 because the ratio of damaged
length to overall rope length is smaller for DPJR2 than it is for DPJR1
as discussed previously. Though the predicted values of the failure
axial strain overestimate the experimentally measured values,
both the test data and simulation results show the same trend—
failure axial strain decreases as the L/d ratio increases. Despite the
fact that the predicted results are significantly greater than the
measured values, these estimates are better than values predicted
by other models previously reported in the research literature
(e.g., net area effect model) that do not show any dependence on
the L/d ratio and do not account for damage. Furthermore, it is
unclear as to how the failure axial strains were measured during
the testing program because, even for the initially undamaged case,
the measured values reported by Ward et al. [18] (approximately
8% for the undamaged case) are below the failure axial strain of
the PET fibers (13%). Considering that the helical nature of the rope
geometry does not have a significant impact on the distribution
of rope component strains because both rope geometries have
parallel sub-rope constructions and the helix angle at the strand
level is less than 10°, the rope failure axial strain is expected to
be larger than what was reported. Hence, the differences between
the measured and predicted values may be attributable to several
factors, and additional research is needed to identify the potential
sources of error.
4. Conclusions
In this paper, a detailed description of a model for predicting
the effect of broken rope components on overall rope response
was presented. The proposed model depends on the ability of
broken rope components to pick up their proportionate share
of axial loads over a distance measured from the failure region,
which is referred to as the recovery length. The model assumes
axisymmetric response and ignores any loss of symmetry due to
the failure of any rope components. If the recovery length value
is admissible, a damaged rope is discretized along its length into
two-noded axial-torsional elements, which provides a means for
1708 J.F. Beltrán, E.B. Williamson / Engineering Structures 33 (2011) 1698–1709
representing a weakened cross-section acting over a localized
region. The developed model takes into account the geometric
nonlinearity of a rope’s geometry and the potential nonlinearity
of the constitutive response of the components that form a rope. In
addition, the potential confinement effect of a rope jacket on rope
components is included in the proposed model by assuming the
rope jacket behaves like a thin-walled tube.
Parametric studies were performed to show how two selected
parameters, the initial loss of rope cross-sectional area and the
L/d ratio, influence the predicted values of axial load capacity and
failure axial strain of two PET ropes. Results showed that rope
capacity decreased as the initial loss of rope cross-sectional area
increased, but it did not show a strong dependence on the L/d ratio
for all the initial damage states considered. The reduction of rope
capacity was mainly proportional to the loss of cross-sectional area
due to rope component failures (net area effect model). Conversely,
failure axial strain decreased and showed a decreasing rate of
change as both the initial loss of rope cross-sectional area and the
L/d ratio increased. Damaged rope stiffness tended to converge to
the initial undamaged rope stiffness as the value of the L/d ratio
became larger, and the rate of convergence increased as the initial
damage state of the rope decreased.
Experimental values obtained from static capacity tests on
different types of PET rope constructions were used to evaluate
the proposed model. Predicted values of the failure axial load
of an initially damaged rope showed good correlation with the
experimental data for the ropes analyzed. Experimental values,
however, did not show a clear trend in their dependency on the
L/d ratio. For the two ropes analyzed in this study, the proposed
numerical model overestimated the rope failure axial strains. The
experimental and predicted values of the rope failure axial strain,
however, showed a similar trend in their dependence on the L/d
ratio: as the value of the L/d ratio increased, the value of failure
axial strain decreased.
The proposed model can be used to perform parametric studies
to evaluate how other variables, such as location of the damaged
cross-section along the rope length, confinement exerted by the
presence of a rope jacket, and rope cross-sectional models (i.e.,
packing or wedging geometry), influence the overall behavior
of damaged ropes. The authors have validated the proposed
model using experimental data obtained from large-scale tests
on PET damaged ropes, and a forthcoming paper will describe
these results. While the current model can be used to obtain a
reasonable estimate of rope behavior, improvements—including
the consideration of the loss of symmetry due to the failure of
individual rope components—are still needed before it can be
reliably used for detailed design calculations. Due to the generality
of its formulation, the proposed model can be extended to consider
multiple failure regions along the length of a given rope, it
can incorporate several axial-torsional elements throughout the
recovery length considering the appropriate stiffness of the rope
components, and it can represent damaged ropes comprised of
materials other than PET.
Appendix
Rope response is computed by summing up the contribution
of each component that forms the rope that is being analyzed.
Based on experimental results [38], PET fibers experience damage
when subjected to stretching. Beltran and Williamson [11]
proposed a model to compute rope response considering the
degradation of its mechanical properties. Thus, the damage-
dependent constitutive relationship that governs the normal stress
for each rope component that belongs to the lowest hierarchical
structure of the rope cross-section modeled, σd (ε, ε 2 , . . .), is
assumed to be a polynomial function of the normal strain up to
the fifth degree and has the following form:

εε 2 ε
    3 
σd (ε, ε , . . .) = (1 − D) a1
2
+ a3+ a2
εb
εb εb
 5 
ε ε
 4
+ a4 + a5 σb (18)
εb εb
where ε is the axial strain experienced by the rope component,
εb is the normal strain at which the rope component reaches its
maximum normal (σb ) stress under monotonic loading, and D is
a damage index that can vary from 0 to 1. The coefficients ai are
constitutive parameters that are chosen to provide a best fit to
measured data for rope components that belong to the lowest
hierarchical level of a rope.
Under the assumption that the strain range experienced by rope
components is the main source of damage under static tension
loading conditions, Beltran and Williamson [17] proposed that the
evolution of the damage index D can be computed as
η−ηu
D(η) = DI + α1 ηβ1 + (1 − α1 ηuβ1 − DI )e θ1 (19)
where the value of DI is the initial damage that a rope component
could have before being loaded, the coefficients α1 and β1 are
constitutive parameters that can be obtained from experimental
data, θ1 is a constant much smaller than 1 (θ1 ≪ 1), and η is
a dimensionless variable defined as (ε − εt )/εb , where εt is the
threshold strain that must be exceeded before damage occurs.
The values of the damage parameters assumed for the analyses
presented in this paper are: α1 = 0.12, β1 = 0.84, εt = 0.04 and
θ1 = 0.001, which come from the results of extensive parameter
studies on small-scale PET ropes [17].
References
[1] Chaplin CR. Torsional failure of wire rope mooring line during installation in
deep water. Eng Fail Anal 1998;6:67–82.
[2] Utting WS, Jones N. A survey of literature on the behaviour of wire ropes. Wire
Ind 1984;51(109):623–9.
[3] Foster GP. Advantages of fiber rope over wire rope. J Ind Text 2002;32:67–75.
[4] Cardou A, Jolicoeur C. Mechanical models of helical strands. Appl Mech Rev
1997;50(1):1–14.
[5] Ghoreishi S, Cartraud P, Davies P, Messager T. Analytical modeling of synthetic
fiber ropes subjected to axial loads. Part I: a new continuum model for
multilayered fibrous structures. Internat J Solids Structures 2007;44:2924–42.
[6] Liu F. Static response model of synthetic fiber lines with internal friction.
Report TM 44-89-14. Port (Hueneme, CA): Naval Civil Engineering Laboratory;
1989.
[7] Liu F. Computer modeling of high-strength synthetic lines under static and
cyclical tensile loads. Technical report TR-2052-OCN. Port (Hueneme, CA):
Naval Facilities Engineering Services; 1995.
[8] Lo KH, Xu H, Skogsberg LA. Polyester rope mooring design considerations.
In: International offshore and polar engineering conference. 1999.
[9] Karayaka M, Srinivasan S, Wang S. Advanced designed methodology for
synthetic mooring. In: Offshore technology conference. OTC 10912. 1999.
[10] Seo MH, Backer S, Mandell JF. Modelling of synthetic fiber ropes deterioration
caused by internal abrasion and tensile fatigue. Washington (DC, USA): MTS
Science and Technology for a New Oceans Decade; 1990.
[11] Beltran JF, Williamson EB. Degradation of rope properties under increasing
monotonic load. Ocean Eng 2005;32:823–44.
[12] Beltran JF, Williamson EB. Numerical simulation of damage localization in
polyester mooring ropes. J Eng Mech 2010;136:945–59.
[13] Costello GA. Theory of wire rope. New York (USA): Springer-Verlag; 1990.
[14] Leech CM. Theory and numerical methods for the modeling of synthetic ropes.
Commun Appl Numer Methods 1987;3:407–713.
[15] Leech CM. The modeling of friction in polymer fibre ropes. Int J Mech Sci 2002;
44:621–43.
[16] Li D, Miyase A, Williams JG, Wang SS. Damage tolerance of synthetic-fiber
mooring ropes: small-scale experiments and analytical evaluation of damaged
subropes and elements. Technical report CEAC-TR-03-0101. University of
Houston; 2002.
[17] Beltran JF, Williamson EB. Investigation of damage-dependent response of
mooring ropes. J Eng Mech 2009;135:1237–47.
 J.F. Beltrán, E.B. Williamson / Engineering Structures 33 (2011) 1698–1709
[18] Ward EG, Ayres RR, Banfield S, O’Hear N, Smith CE. Experimental investigation
of damage tolerance of polyester ropes. In: Proceedings of the 4th international
conference on composite materials for offshore operations. 2005.
[19] Hankus J. Safety factor for hosting rope weakened by fatigue cracks in wires.
In: Organisation internationale pour l’Etude de l’Endurance des cables. OIPEEC.
Round table conference. 1981.
[20] Chaplin CR, Tantrum N. The influence of wire break distribution on strength.
In: Organisation internationale pour l’Etude de l’Endurance des cables. OIPEEC.
Round table conference. 1985.
[21] Cholewa W, Hansel J. The influence of the distribution of wire rope faults
on the actual breaking load. In: Organisation internationale pour l’Etude de
l’Endurance des cables. OIPEEC. Round table conference. 1981.
[22] Cholewa W. Wire fracture and weakening of wire ropes. Wire rope discard
criteria. Round table conference. Swiss federal institute of technology.
ETH. Institute of lightweight structures and ropeways. September. Zurich
(Switzerland); 1989.
[23] Evans JJ, Ridge IML, Chaplin CR. Wire failures in ropes and their influence
on local wire strain behaviour in tension–tension fatigue. J Strain Anal 2001;
36(2):231–44.
[24] Chien CH, Costello GA. Effective length of a fractured wire in wire rope. J Eng
Mech 1985;111(7):952–61.
[25] Raoof M. Wire recovery length in a helical strand under axial-fatigue loading.
Int J Fatigue 1991;13(2):127–32.
[26] Raoof M, Huang YP. Wire recovery length in suspension bridge cable. J Struct
Eng 1992;118(12):3255–67.
[27] Raoof M, Kraincanic I. Recovery length in multilayered spiral strands. J Eng
Mech 1995;121(7):795–800.
1709
[28] Raoof M, Kraincanic I. Determination of wire recovery length in steel cables
and its practical applications. Comput Struct 1998;68:445–59.
[29] Winkler MM, McKenna HA. The polyester rope taut leg mooring concept:
a feasible means for reducing deepwater mooring cost and improving
stationkeeping performance. In: 27th Annual offshore technology conference.
1995.
[30] Gjelsvik A. Development length for single wire in suspension bridge cable.
J Struct Eng 1991;117(4):1189–201.
[31] LeClair R. Axial response of multilayered strands with compliant layers. J Eng
Mech 1991;117(12):2884–903.
[32] Phillips JW, Miller RE, Costello GA. Contact stresses in a straight cross-lay
wire rope. In: Proc., 1st annual wire rope conf. Engineering extension service.
Washington State University. 1980.
[33] Beltran JF. Computational modeling of synthetic–fiber ropes. Ph.D. disserta-
tion. Austin (TX): The University of Texas at Austin; 2006.
[34] Banfield S. Personal communication.
[35] Ayres R. Determining interim criteria for replacing damaged polyester
mooring rope. Final report to BP exploration. PN 6941-RRA. 2001.
[36] Leech CM, Hearle JWS, Overington MS, Banfield SJ. Modelling tension and
torque properties of fibre ropes and splices. In: Proceedings of the third
international offshore and polar engineering conference. 1993.
[37] Banfield SJ, Hearle JWS, Leech CM, Tebay R, Lawrence CA. Fibre rope modeller
(FRM): a cad program for the performance and prediction of advanced cords
and ropes under complex loading environments. Germany: Techtextil; 2001.
[38] Van den Heuvel CJM, Heuvel HM, Faasen WA, Veurink J, Lucas LJ. Molecular
changes of PET yarns during stretching measured with rheo-optical infrared
spectroscopy and other techniques. J Appl Polym Sci 1993;49:925–34.