In this section, results are presented first for a single cable using the full-geometry and efficient modeling techniques. Afterward, results of three-cable specimen models are presented. The cable-modeling approaches are evaluated using the three-cable models in terms of the specimen’s stiffness, deformation field, and strain fields in the rubber.
3.1. Full-Geometry Single-Cable Model
The full-geometry model of the seven-wire strand is used to obtain the components of the stiffness matrix that are used to evaluate the efficient cable models. Here, we look into the nonlinearity of the overall stiffness response of the full model and the influence of element type, mesh size, and cable length on the cable‘s stiffness, which is relevant for the full-geometry three-cable specimen of
Section 3.3. The length of the modeled cable is quantified as a fraction of the strand length (the axial distance so that the outer wires are completely wound around the cable axis, as described in more detail in
Section 2.3.1) and denoted as relative cable length.
To show the nonlinearity in the cable stiffness, two load cases are studied. A cable load (
,
) of (5%, 0,0) and (
10/m) is applied in load case A and load case B, respectively.
Figure 6 shows the axial force
, the torsional moment
and the bending moment
of the cable over the longitudinal strain
(load case A,
Figure 6a) and over the curvature
(load case B,
Figure 6b). For a mesh size of 0.5 mm and a relative cable length of 0.8, the
and
plots are approximately linear, whereas the
curve shows a slight nonlinearity towards higher curvatures.
The influence of cable length and mesh size are investigated for applied loads of
= 0.5%,
= 2 rad/m, and
= 1/m, which are applied individually. The longitudinal strain of 0.5% corresponds to a maximum Mises stress of
= 960 MPa in the central wire and a total force of
= 60 kN. We here assume that those loads cover the relevant range for the intended applications and that nonlinear effects that occur at higher loads can be neglected. The stiffness parameters are evaluated as secant stiffnesses of the loading curves and are plotted over the cable length and mesh size in
Figure 7. The range and units of the individual stiffness parameters
,
,
,
, and
are quite different. To better visualize the dependency of those parameters on the cable length and mesh size, they are plotted relative to their respective most accurate values (such as those obtained for either highest cable length or smallest mesh size, as explained in the following).
In
Figure 7a, the cable length is varied for linear hexahedral elements with reduced integration and a fixed mesh size of 0.5 mm. The relative cable length of 1.4 is assumed to give the most accurate results, so those
components are used to normalize the respective results of the other models. As expected, the rigid connection from the cable ends to their corresponding reference points introduces numerical artifacts that increase the evaluated
components for smaller cable lengths. The bending stiffness
is particularly sensitive to these cable end effects.
The element type and mesh size are varied in
Figure 7b for a constant relative cable length of 1.2, since this is the length for which the stiffness parameters have already reached a plateau, as shown in
Figure 7a. The results for the smallest mesh size of the quadratic tetrahedral elements (0.75 mm) are used to normalize
. The curves for bilinear hexahedral and quadratic tetrahedral elements show that the finer the mesh size, the higher the computed stiffness components. For the hexahedral elements, no clear plateau of
components is reached for the finest mesh size of 0.3 mm. This indicates that bilinear hexahedral elements would need to be much finer to accurately compute the cable’s stiffness. The quadratic tetrahedral element results show a plateau at a mesh size of about 1 mm. Similar to the cable length study, the bending stiffness
is most sensitive to the mesh size. The 1.25 mm mesh with quadratic tetrahedral elements (see the pictogram in
Figure 7b) yields acceptable computation times and quite accurate results: The stiffness parameters are up to 4% lower than for a mesh size of 0.75 mm. Therefore, in the bigger three-cable specimen models with full geometry, quadratic tetrahedral elements with a mesh size of 1.25 mm are used.
Table 4 lists the model size, necessary RAM, and computation time for the full-geometry single-cable models of
Figure 7. To keep the table short, only model parameters of the maximum and minimum cable length (length study) and mesh size (mesh study) are listed. For bilinear hexahedral elements, no mesh convergence is reached at a mesh size of 0.3 mm with computation times of 20 min. The finest quadratic tetrahedral element results took about 5 min to compute.
The finest mesh size of the tetrahedral elements is assumed overall to give sufficiently accurate
components. Those
components for the quadratic tetrahedral elements with a mesh size of 0.75 mm and a relative cable length of 1.2 are therefore extracted. The efficient cable models are set up to fit these components:
Since the matrix is nearly symmetrical, we make it symmetrical by setting
and
, which are much smaller than the other components, to zero and introduce a parameter
that we use in the following for both
and
:
This simplification results in only four
parameters that should be reached in the efficient cable models; see
Table 5. Inverting the simplified
matrix yields
. As mentioned in
Section 2.1, the cubic modeling approaches can be fitted based on either
or
.
3.2. Efficient Single-Cable Models
The cable-modeling approaches introduced in
Section 2.2 are set up as described in
Section 2.3.
Table 6 lists the parameters that are either calculated analytically or calibrated using the cable FEM models. To obtain the model parameters, the stiffness parameter
or
are used. For the approaches that do not have a tension/torsion coupling, the parameters are calculated once with
and once with
.
Figure 8 shows the components of
and
obtained for the cubic modeling approaches. The diagrams use a logarithmic scale with relative values normalized to the target
or
values stated in
Table 5.
Figure 8a,b show the stiffness values for cable model parameters calculated to fit
. As expected, the stiffness parameters obtained for
plotted in
Figure 8a fit well to the target values, except for the bending stiffness in the solid approach, which is too high by a factor of 9. The fit of the beam and solid/beam approaches is equally good. The
components for the same efficient cable models, however, are about 53% higher than the components of the full-geometry model. When the cable models are calibrated to
(see
Figure 8c,d), the
components fit well, but components of
are lower by about 36%. This shows that for a cable that has tension/torsion coupling (
), an efficient cable model with cubic material can fit either
or
but not both at the same time.
corresponds to the stiffness in tension with constrained torsion and
to tension with free torsion. When using such a cable model with cubic material, the model’s parameters should be calculated depending on the application of the model. If the application is unknown, an intermediate stiffness of
and
should be used for the models.
Three efficient cable-modeling approaches that can account for the tension/torsion coupling are investigated, and their
and
components are plotted in
Figure 9. Similar to the cubic approaches, the bending stiffness of the solid approach is too high by a factor of 7.8. The solid, solid–HGO, and beam approaches can capture
well; see
Figure 9a. The largest differences are observed for
in the solid–HGO approach, which is 22% higher than the target value. Note that due to having only three calibration parameters in the HGO approach, the four independent stiffness parameters cannot all be fitted at the same time. Other HGO parameters such as
D,
, and
could be fitted as well but do not help to improve the accuracy of
.
The components of the
matrix of the solid and the beam approach in
Figure 9b fit well to the target
components, except for the
of the linear elastic solid approach. For the solid–HGO approach, only the
component fits well, whereas the other
terms are lower by 46% to 60%. This is due to an amplification of the deviation of the
components since the
matrix is inverted to calculate
. Furthermore, the convergence in the simulations with the solid HGO approach is bad, which requires many more iterations in the FEM simulation. The solid approach can therefore be used for applications where bending does not play a role, and the beam approach can be used in all cases where inaccuracies related to the coupling of the beam nodes to the rubber are acceptable—for example, because the rubber/cable interface is not of special interest in the model.