+ MODEL

Available online at www.sciencedirect.com

ScienceDirect
Publishing Services by Elsevier
International Journal of Naval Architecture and Ocean Engineering xx (2016) 1e10
http://www.journals.elsevier.com/international-journal-of-naval-architecture-and-ocean-engineering/

Peridynamic simulation of brittle-ice crushed by a vertical structure

Minghao Liu, Qing Wang*, Wei Lu
College of Ship Engineering, Harbin Engineering University, Harbin 150001, China
Received 25 June 2016; revised 4 September 2016; accepted 18 October 2016
Available online ▪ ▪ ▪

Abstract

Sea ice is the main factor affecting the safety of the Arctic engineering. However, traditional numerical methods derived from classical
continuum mechanics have difficulties in resolving discontinuous problems like ice damage. In this paper, a non-local, meshfree numerical
method called “peridynamics”, which is based on integral form, was applied to simulate the interaction between level ice and a cylindrical,
vertical, rigid structure at different velocities. Ice in the simulation was freshwater ice and simplified as elastic-brittle material with a linear
elastic constitutive model and critical equivalent strain criterion for material failure in state-based peridynamics. The ice forces obtained from
peridynamic simulation are in the same order as experimental data. Numerical visualization shows advantages of applying peridynamics on ice
damage. To study the repetitive nature of ice force, damage zone lengths of crushing failure were computed and conclude that damage zone
lengths are 0.15e0.2 times as ice thickness.
Copyright © 2016 Society of Naval Architects of Korea. Production and hosting by Elsevier B.V. This is an open access article under the
CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords: Numerical simulation; Iceestructure interaction; State-based peridynamics; Ice crushing; Ice forces

1. Introduction conducted by Tian and Huang (2012) to verify the lower ice
force on the structure, and the experimental flexural strength
With increasing interest in oil and gas exploration in cold results were discussed. From 2004 to 2005, Huang et al.
regions, human activities in these areas are increasing. In (2007) conducted series of model tests to explore the mech-
Arctic region, the presence of sea ice is an important factor anism that controls the procedure of ice induced vibration of
that contributes to the performance of naval architecture. For vertical narrow piles and setup interaction coefficient that is
this reason, much research has focused on cold-region engi- able to reflect the interaction level of ice and structure. Yue
neering. Vertical structure is a typical structural form in Arctic et al. (2009) conducted full-scale tests of interaction between
ocean engineering, such as the jacket platforms being widely level ice and vertical compliant structures to study dynamic
used in Bohai Sea. Ice induced force during interaction of ice ice forces and structure vibrations. Regarding with loading
and vertical structures attracts great attention from scholars. rate, ductile, ductileebrittle transition and brittle ice failure
Sodhi and Morris (1984) conducted small-scale experi- were observed.
ments to measure varied ice forces by pushing rigid cylindrical Ice damage during iceestructure interaction is a complex
structures of different diameters at different velocities through process of material failure. Numerical techniques for
a level ice sheet. The small-scale experiments showed that the addressing discontinuities are of significant interest to re-
characteristic frequency maintains a linear relationship with searchers. But currently, it is short of a numerical method to
the ratio of velocity to thickness. Scale-model tests of in- simulate the complex failure process of ice (Bergan et al.,
teractions between level ice and conical structures were 2010). For the macro approach based on continuum me-
chanics, the solution of governing equation is related to the
partial derivatives of its stress and relative displacement.
* Corresponding author. However, the numerical methods derived from classical
E-mail address: [email protected] (Q. Wang).

http://dx.doi.org/10.1016/j.ijnaoe.2016.10.003
2092-6782/Copyright © 2016 Society of Naval Architects of Korea. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Please cite this article in press as: Liu, M., et al., Peridynamic simulation of brittle-ice crushed by a vertical structure, International Journal of Naval Architecture
and Ocean Engineering (2016), http://dx.doi.org/10.1016/j.ijnaoe.2016.10.003
 + MODEL

2 M. Liu et al. / International Journal of Naval Architecture and Ocean Engineering xx (2016) 1e10

continuum mechanics, which has difficulties in resolving structure at different velocities. The basic theory of state-based
discontinuity, have also restricted the research progress of the peridynamics and numerical methods as implemented in the
polar ocean engineering. Peridynamics is a meshless method simulation, such as discretization and time-integration stabil-
to solve the equation of motion by using integral form. ity, were introduced briefly. To implement a two-body inter-
Therefore, it has obvious advantages for the simulation of action in the peridynamic model, an interaction force density
complex material damage. in terms of relative positions in the current configuration was
Silling (2000) proposed the original peridynamic theory applied to prevent material particles that belong to different
called bond-based peridynamics in 2000, and it reformulates bodies from sharing the same position. To verify the feasibility
the continuum mechanics equations. Similar to molecular and accuracy of applying the peridynamic method to ice-
dynamics, peridynamic equation of motion uses a nonlocal estructure interactions, the peridynamic simulation results
method to describe the force between particles, so that there is were compared with small-scale experiments. The results from
no requirement to know the crack location in advance. The numerical simulation, including the ice force values, repetitive
essence of the equations is that integration rather than differ- nature of ice forces, phenomenon of damaged ice, damage
entiation used in the calculation, which also avoids redefining zone length and characteristics of ice loading cycle were
the body of discontinuity (see Fig. 1). Brbaru et al. (2009) presented and analyzed.
developed adaptive refinement algorithms for non-local peri-
dynamics in elastic material regarding with one dimension 2. Theory of peridynamics
model. In addition, compared with classical elasticity, they
discussed many types of numerical convergence for peridy- The peridynamic theory divides continuum into discrete
namics and obtained uniform convergence to the classical material points, see Fig. 3. Therefore, the deformation of body
solutions of static and dynamic problems in one dimension in can be described by displacements of points. Each material
the limit of the horizon going to zero. point interacts with its neighbors by nonlocal forces, see
Even though bond-based peridynamics shows remarkable Fig. 2. Integrating the forces acting on a material point leads to
ability to deal with discontinuities in material like crack its acceleration and velocity with help of Newton's law and
propagation, fracture and damage, its weak relationship with avoids the mathematical problems faced in partial differential
classical mechanics restricts its application on complicated methods when addressing discontinuities (Silling, 2000).
material. As a result, Silling et al. (2007) proposed a more Furthermore, when the relative displacement of material
general form of peridynamics, called state-based peridy- points reaches a critical value, nonlocal force vanishes. This
namics, which can adopt the constitutive model directly and process means appearance of material failure. Additional
reinforce the relationship between peridynamics and classical criteria are not required for determining crack growth and
mechanics. Warren et al. (2009) applied state-based peridy- branching, which allows crack growth and branching to pro-
namics to solve deformation and fracture problems in solid ceed spontaneously.
materials. Foster et al. (2010) proposed a method to implement Peridynamics introduces a “horizon” for each material
a rate-dependent viscoplastic material within this peridynamic point, in which the material points interact with each other by
model.
The peridynamic theory overcomes weaknesses of existing
methods and can simultaneously and spontaneously simulate
all failure modes without simplifying assumptions. The effi-
ciency and convenience of this numerical method facilitate its
application in deformation and discontinuities of materials.
In this paper, we applied state-based peridynamics, which is
a nonlocal theory that uses integration rather than differenti-
ation to calculate material deformation and requires no need
for a predefined crack propagation path, to simulate the
interaction between level ice and a cylindrical, vertical, rigid

Fig. 2. Each point x in the body interacts with points in its horizon, d is radius
Fig. 1. Comparison of local method, peridynamics and molecular dynamics. of horizon.

Please cite this article in press as: Liu, M., et al., Peridynamic simulation of brittle-ice crushed by a vertical structure, International Journal of Naval Architecture
and Ocean Engineering (2016), http://dx.doi.org/10.1016/j.ijnaoe.2016.10.003
 + MODEL

M. Liu et al. / International Journal of Naval Architecture and Ocean Engineering xx (2016) 1e10 3

displacement, where u½x; t and u½x0 ; t are displacements of

points initially at x and x′, respectively. The deformation-
vector state is defined as
Y½x; t〈x〉 ¼ x þ h ð2Þ
To associate peridynamics with classical mechanics and to
introduce a constitutive model into the theory, a non-local
shape tensor for a point x is defined as
Z
K½x; t ¼ uðjxjÞðx5xÞdVx0 ð3Þ
Hx

Fig. 3. Configuration of continuum discrete in peridynamics.

where 5 is the dyadic product operator and u is an scalar
influence function that is dependent on jxj. Influence function
“bonds” and forces are integrated over the horizon. As a result, u is a parameter whose value is 0  u  1 and describes the
a material point is subjected to forces from not only the nearest effect degree between a pair of material points. Traditionally,
points but all points in the horizon, see Fig. 2. It is for this the nearer the two points are, u is closer to 1. The value of
reason that peridynamics is known as a nonlocal method. The parameter u needs more investigation. In this paper, we only
points in the horizon of point x are called family points of make u equal to 0 or 1 and u ¼ 0 means bond breakage.
point x. With the shape tensor, the non-local deformation gradient F
The original peridynamic theory, called bond-based peri- is given by
dynamics, is a convenient and efficient numerical method to "Z #
deal with discontinuous problems. However, this original
bond-based peridynamics is an oversimplification with the F½x; t ¼ uðjxjÞðY〈x〉5xÞdVx0 K1 ð4Þ
Hx
assumption that any pair of material particles interacts with
each other only through a central force potential without where Hx ¼ fjx0  xj < dg.
dependence on other local conditions (Foster et al., 2010). To associate force state with material deformation, tradi-
This leads to some critical shortcomings, such as an effective tional stress-strain model from classical mechanic theory can
Poisson's ratio of 1/4 and a limited relationship with conven- be used as an intermediate step. After that, Cauchy stress
tional mechanics. To address these issues, a more general tensor can be described as a function of deformation gradient,
theory termed state-based peridynamics was proposed by sðFÞ. We can now convert Cauchy stress into first Pio-
Silling et al. (2007) and be achieved using the idea of peri- laeKirchhoff stress tensor
dynamic states. This theory reinforces the relations between
peridynamics and classical mechanics and offers an effective P ¼ detðFÞsFT ð5Þ
method to adopt constitutive models. The force-vector state can be written as
We now review the fundamental theory of state-based
peridynamic theory briefly. In state-based peridynamics, T〈x〉 ¼ uðjxjÞPðFÞK1 x ð6Þ
deformation state Y and force state T vectors are defined to
associate a vector with each bond x ¼ x  x0 within a neigh- where PðFÞ is the first Piola-Kirchhoff stress tensor obtained
borhood of any point x. The bond that the vector state operates from the constitutive model as a function of deformation
on is written in angle brackets. For state-based peridynamics, gradient. The force state vector can be substituted back into
the equation of motion can be written as Eq. (1) and to find accelerations at this step. The process of
integration is purely meshless Lagrangian at each material
Z
point without need for Gauss integration points and interpo-
r€
u½x; t ¼ fT½x; t〈x0  x〉  T½x0 ; t〈x  x0 〉gdVx0 þ b½x; t lation of nodal value.
Hx

ð1Þ 3. Constitutive model and failure criterion

where T½x; t is the force-vector state that represents the Ice is a complicated material that is composed of solid ice,
relationship (bond force) between material points at time t, brine, gas and depends on temperature, various types of solid
and maps the deformation-vector state into a force-vector state salts. Environmentally controlled variations in the mechanisms
for each material point. Hx is the space of material points that of ice growth can result in several different grain structures,
neighbor point x, which means the area of horizon. r is the depending on the prevailing conditions. The most common
material density and b½x; t is the body force that acts on point grain structures are granular, columnar and discontinuous
x. For convenience, in the following discussion, we denote x ¼ columnar. Granular ice is usually isotropic (ice properties do
x  x0 as the relative position of two points in reference not depend on the direction of measurement). Furthermore,
configuration and h ¼ u½x0 ; t  u½x; t as their relative granular ice can be produced conveniently in indoor

Please cite this article in press as: Liu, M., et al., Peridynamic simulation of brittle-ice crushed by a vertical structure, International Journal of Naval Architecture
and Ocean Engineering (2016), http://dx.doi.org/10.1016/j.ijnaoe.2016.10.003
 + MODEL

4 M. Liu et al. / International Journal of Naval Architecture and Ocean Engineering xx (2016) 1e10

laboratories. Thus, granular ice is a frequently used material in critical stretch used in bond-based peridynamics (Silling,
laboratory tests and numerical simulations. 2000), which is a typical method to describe failure in brit-
Furthermore, since stress state during ice crushing is mostly tle materials.
compressive, ductileebrittle transition of ice compressive
strength should be a complex factor (Jones, 1982). According 4. Numerical methods
to Yue et al. (2009), during ice crushing, ice forces represent
three modes regarding with loading rate. Peridynamics is typical meshfree numerical method to
simulate material fracture. In this section, several numerical
(1) When loading rate is lower than 20 mm/s, ice force is methods, including discretization of equation of motion, time
quasi-static. Ice is ductile material with plastic mechanical integration and numerical stability, are presented for imple-
behavior. mentation of peridynamics.
(2) When loading rate is between 20 mm/s to 40 mm/s, ice
force is defined as locked-in ice force. Ice behaves duc- 4.1. Discretization
tileebrittle transition.
(3) When loading rate is higher than 40 mm/s, ice force be- After the continuum domain has been discretized by unit
comes irregular. At this state, ice is elastic brittle material. cell volumes occupied by material points, Eq. (1) can be
transformed to the discretized form
Several researchers have conducted laboratory tests to
X
m
examine brittle strength, failure patterns and characteristics of rðxi Þ€
uðxi ; tÞ ¼ fT½xi ; t〈xn  xi 〉  T½xn ; t〈xi  xn 〉gVn
iceestructure interactions (Kuehn et al., 1993; Kamio et al., n¼1
2003; Schulson, 1997, 2001; Jordaan, 2001). In addition, the
þ bðxi ; tÞ
moving velocities of vertical structure in present simulation
are higher than 50 mm/s, which means the ice surrounding the ð10Þ
structure, behaves mainly elastic brittle ice. Since brittle ice
failure plays a critical role in ice failure modes surrounding where T is the force state function, point i is the subjective
naval structures, a linear elastic constitutive model is used in point in a given horizon and n is family point in horizon. m is
this paper with a critical equivalent strain criterion to describe the number of points that interact with point i by unbroken
elastic brittle ice. bonds. The unit volume is Vn ¼ ðDxÞ3 for a three-dimensional
As described in Section 2, with non-local deformation model and Vn ¼ ðDxÞ2 for a two-dimensional model (Dx is the
gradient F, strain of a material point can be obtained. The point spacing).
strain measurement from simulation was calculated by using Similarly, the shape tensor and deformation gradient can be
Green strain as shown in the following equation discretized as

1 X
m

E ¼ ðFT F  IÞ ð7Þ Kðxi Þ ¼ uðjxn  xi jÞððxn  xi Þ5ðxn  xi ÞÞVn ð11Þ

2 n¼1

where I is unit tensor. and

For elastic brittle ice in simulation, constitutive model can " #
be numerically updated as X
m
Fðxi Þ ¼ uðjxn  xi jÞðY〈xn  xi 〉5ðxn  xi ÞÞVn K1
s¼C:E ð8Þ n¼1

ð12Þ
where C is elastic material tensor.
In peridynamics, various methods exist to determine
criteria of material damage. We use a critical equivalent strain 4.2. Time integration and numerical stability
to determine breakage of the bond between x and x0 in elastic
brittle freshwater ice in which shear strain plays an important To acquire displacements of points, an explicit method is
role in material failure. Firstly, we average the values of strains applied to the preceding time integration. According to Eq.
at x and x0 as Eðx; x0 Þ ¼ ðEðxÞ þ Eðx0 ÞÞ=2. The equivalent (1), acceleration of point i at time step n can be calculated by
strain can be expressed as
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi f ni þ bni
€ni ¼
u ð13Þ
3 ri
Eeq ðx; x0 Þ ¼ EIJ ðx; x0 Þ : EIJ ðx; x0 Þ ð9Þ
2
where f is integral non-local force vector exerting on point i.
where EIJ is the deviatoric part of the averaged strain. After accelerations of material points is acquired, explicit al-
If the value of the equivalent strain becomes greater than a gorithm can be used to calculate the velocities and displace-
critical value, the bond breaks irreversibly. This is similar to ments at time t

Please cite this article in press as: Liu, M., et al., Peridynamic simulation of brittle-ice crushed by a vertical structure, International Journal of Naval Architecture
and Ocean Engineering (2016), http://dx.doi.org/10.1016/j.ijnaoe.2016.10.003
 + MODEL

M. Liu et al. / International Journal of Naval Architecture and Ocean Engineering xx (2016) 1e10 5

u_ ni ¼ u_ n1
i €ni $Dt
þu ð14Þ Fortran codes and compiled in Intel Visual Fortran platform.
Required data and visualization are plotted in Tecplot 360
uni ¼ un1
i þ u_ ni $Dt ð15Þ software. To verify the feasibility of this numerical method,
ice forces of three operating conditions with different moving
Although the time integration method is straightforward,
velocities were compared with experimental data.
the integration process is conditionally stable. Strict require-
ment of time intervals should be achieved for explicit algo-
rithm. Silling and Askari (2005) applied the von Neumann 5.1. Problem setup
stability analysis to limit the time step size Dt, which opti-
Numerical examples were conducted by pushing a rigid
mizes the stability of the explicit time integration
cylindrical structure through level ice with the peridynamic
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2r method. The pile diameter and length were d ¼ 50 mm and
Dt < P ð16Þ l ¼ 500 mm, respectively. The pile that impacts the ice sheet
VCip
moves with various velocities from 30 mm/s to 1 m/s. Because
the pile vibration is not considered, the pile modeled numer-
where r is material density, V unit volume occupied by ma-
ically by peridynamics is a rigid body that is free from
terial point and Cip ¼ vf=vh.
deformation. Level ice is a rectangular ice sheet and its length,
width and thickness as impacted by the pile are L ¼ 1 m,
4.3. Interaction force
W ¼ 1 m and T ¼ 0.06 m, respectively. The edges of the ice
sheet, except the one interacting with the pile, are fixed. The
In the case of interact problem, a force between material
configuration of the numerical setup is shown in Fig. 4. The
points belong to different bodies is defined by Madenci and
material properties of the ice plate are given as an elastic
Oterkus (2014). This kind of force is based on a critical dis-
modulus of E ¼ 120 Mpa, mass density of r ¼ 900 kg=m3 and
tance, rsh, between two material points of two different bodies
Poisson's ratio of n ¼ 0:3.
when they come close to each other. Since specific two points
For peridynamic method, balance of sufficient accuracy
is close enough, they begin to push each other in order to
and numerical efficiency regarding with number of grid
prevent sharing a same location. The force for interaction is
points is of great importance. According to Madenci and
expressed as
Oterkus (2014), grid size convergence was analyzed consid-
8   !9 ering the vibration of bar. To visualize the effect of grid size
  yj  yn < yj  yn  =
f sh yj ; yn ¼   min 0; csh  1 ð17Þ on numerical error, numerical models with four grid sizes,
yj  yn  : 2rsh ; Dx ¼ L=10; L=100; L=1000 and 10000=L, are developed. The
results show that the error of coarse grid Dx ¼ L=10 increases
where the force constant, csh, can be chosen as with simulation process and the other conditions meet the
requirement of accuracy. Furthermore, convergence analysis
csh ¼ 5c ð18Þ involving with horizon size recommends a grid size of
d ¼ 3,Dx to reduce computational time and dependence on
where c is called micro-modulus in bond-based peridynamics. crack propagation. In this paper, a grid spacing of 10 mm was
For 3-D peridynamic simulation, c is expressed as used in the peridynamic model, which means that the ice
18k plate includes approximately 60,000 material points. Horizon
c¼ ð19Þ radius is d ¼ 3,Dx. According to the principle of time sta-
pd4
bility, the size of the time step is 4.66  106 s. The critical
where k is bulk modulus of material. equivalent strain, Ecr, is 0.0325. The boundary region of the
And critical distance, rsh, can be chosen as ice plate contains three layers of material points on each edge
which are fixed and free of deformation. In the simulation,
Dx six pile velocities were used: v ¼ 50 mm/s, v ¼ 130 mm/s,
rsh ¼ ð20Þ
2
Eq. (17) was proposed in EMU manual which is a
computational platform for peridynamics developed by Sandia
National Laboratory. It was originally proposed for bond-
based peridynamics. However, this method is able to repre-
sent good results in present simulation.

5. Numerical simulation

This section presents the process of developing a numerical

simulation, including the operating conditions and numerical
parameters, and the numerical results obtained from the Fig. 4. Schematic diagram of interaction between level ice and cylindrical
simulation. Peridynamic simulations are implemented by structure.

Please cite this article in press as: Liu, M., et al., Peridynamic simulation of brittle-ice crushed by a vertical structure, International Journal of Naval Architecture
and Ocean Engineering (2016), http://dx.doi.org/10.1016/j.ijnaoe.2016.10.003
 + MODEL

6 M. Liu et al. / International Journal of Naval Architecture and Ocean Engineering xx (2016) 1e10

v ¼ 210 mm/s, v ¼ 500 mm/s, v ¼ 750 mm/s and Because only a linear elastic-brittle constitutive is used in
v ¼ 1000 mm/s. Among these operating conditions, the numerical simulation. Furthermore, urea ice is used in the
output of the first three velocities were extracted for com- scale test and the influence of urea on mechanical properties is
parison with the experimental results. undetermined. For numerical simulation, constitutive models
involved with strain rate mostly need empirical parameters,
5.2. Numerical results which is of difficulty to apply urea ice in simulation. The

Figs. 5e10 represent the total ice forces during contact of

the cylindrical structure with level ice computed by peridy-
namics. The results are filtered using a 150-Hz low-pass filter.
Similar to experimental ice-force curves obtained by Sodhi
and Morris (1986), the total ice forces obtained from numer-
ical simulation represent a typical repetitive nature and fre-
quencies of the ice forces increase with increasing cylindrical
structure velocity. In addition, each periodical involves at least
two typical peaks, one with a higher value and the others with
less. This phenomenon is also visible in the ice force figure
with a pile velocity of 50 mm/s.
The stress state of ice sheet near the vertical structure is
compression. As we know, the mechanism behavior of ice
under compression is really complex. According to Yue et al.
(2009), since compressive strength of ice is sensitive to strain
Fig. 6. Ice force from simulation (v ¼ 130 mm/s).
rate, ice crushing force can be divided into three modes: 1)
Ductile behavior under low strain rate. The ice in this state
undergoes plastic deformation and few cracks can be
observed. 2) Brittle behavior under high strain rate. Ice fails
with several major cracks and little deformation can be
observed. 3) Ductile-brittle transition under medium strain
rate. Micro cracks form with no propagation. Compressive
strength of ice varies in different states.
Table 1 shows the mean and maximum total forces during
contact of the cylindrical structure with level ice. By
comparing the ice forces from simulation and experiment,
respectively, we can see the values of peak forces and mean
forces of simulation are in the same order as experimental
data. However, the peak forces from the experimental results
change with cylindrical structure velocity. The reason for this
problem, perhaps, is that the ductile-brittle transition is caused
by different loading rates. In contrast, in this paper, ice forces
obtained by peridynamics are independent of strain rate. Fig. 7. Ice force from simulation (v ¼ 210 mm/s).

Fig. 5. Ice force from simulation (v ¼ 50 mm/s). Fig. 8. Ice force from simulation (v ¼ 500 mm/s).

Please cite this article in press as: Liu, M., et al., Peridynamic simulation of brittle-ice crushed by a vertical structure, International Journal of Naval Architecture
and Ocean Engineering (2016), http://dx.doi.org/10.1016/j.ijnaoe.2016.10.003
 + MODEL

M. Liu et al. / International Journal of Naval Architecture and Ocean Engineering xx (2016) 1e10 7

Fig. 9. Ice force from simulation (v ¼ 750 mm/s).

Fig. 11. Step 10,000 of simulation (v ¼ 750 mm/s).

Fig. 10. Ice force from simulation (v ¼ 1000 mm/s).

Fig. 12. Step 26,000 of simulation (v ¼ 750 mm/s).

Table 1
Comparison of ice force (simulation vs. experiment).
Velocity (mm/s) Simulation Experiment
Mean force (kN) 50 1.3848 5.3859
130 1.3739 1.7309
210 1.4324 1.9136
Peak force (kN) 50 4.4379 13.0527
130 4.6342 5.2479
210 4.9742 3.8522

numerical computation considering strain rate of ice will be

analyzed in future research.
The results are based on a simulation with the peridynamic
method and data recorded during the interaction. The simu-
lation only shows failure of ice crushing during the interaction
Fig. 13. Step 40,000 of simulation (v ¼ 750 mm/s).
between the cylinder structure and the level ice. Figs. 11e14
shows peridynamic visualization of a pile moving through
an ice sheet at 750 mm/s. pile. This phenomenon is also captured by experimental re-
The simulation images show the progress of ice crushing cord. Fig. 16 is an experimental picture of brittle ice failure
failure, and color differences stand for the equivalent stress induced by vertical structure crushing ice sheet. In addition, in
distribution. A high-stress field is initiated around the cylin- the path cleared by cylindrical structures in peridynamic
drical structure and spreads to most areas of the ice sheet. An simulation, line-like hot spot can be observed, which means
accumulation of damaged ice particles is visible in front of the high stress area arranging ice thickness and occupying one

Please cite this article in press as: Liu, M., et al., Peridynamic simulation of brittle-ice crushed by a vertical structure, International Journal of Naval Architecture
and Ocean Engineering (2016), http://dx.doi.org/10.1016/j.ijnaoe.2016.10.003
 + MODEL

8 M. Liu et al. / International Journal of Naval Architecture and Ocean Engineering xx (2016) 1e10

Morphology of ice crushing failure in peridynamic simu-

lation matches the experimental phenomenon proposed by
Sodhi and Morris (1984) and Huang et al. (2007). Simulated
ice failure is mainly caused by spalling and extrusion and
obvious crack involving splitting is not observed. In some
experiments, large crack occasionally happened along with
crushing path. This discrepancy perhaps is induced by high
confinements and grid size. High confinements caused by
boundary conditions may prevent the crack propagation in ice.
At the same time, for peridynamics, material failure happens
spontaneously and identification of specific components of
failure needs more work to be achieved.

5.3. Discussion
Fig. 14. Step 56,000 of simulation (v ¼ 750 mm/s).
The ice forces show secondary peak forces accompanied by
true peak values, especially for the 50 mm/s velocity in ex-
periments. Fortunately, this phenomenon was captured in the
computational simulation by peridynamics.
According to Sodhi and Morris (1986), the repetitive nature
of the ice forces can be characterized by determining the
dominant frequency at which the ice forces vary. This is
defined as the “characteristic frequency of ice failure”. To
determine the characteristic frequency, force data for each
velocity of pile were transformed into a frequency domain by
the Fast-Fourier-Transform (FFT) routine. Figs. 17e22 show
the results of the transformation. It is obvious that the fre-
quency of ice failure increases with increasing velocity of the
moving structure. In addition, variations in increasing the
frequency maintain a slightly increasing trend.
During each loading cycle, the iceestructure interaction
undergoes three stages: interaction, crushing and failure. At
Fig. 15. Local damage in PD simulation. the beginning of the iceestructure interaction, the ice force
appears to be increasing. Thereafter, as the pile continues to
crush the ice sheet, the ice force increases with crushing dis-
tance. Finally, ice failure emerges when the crushing distance
reaches a critical value. To quantify the distance the ice or
structure moves for each loading cycle, a term “damage zone”
was proposed by Sodhi and Morris (1986). The distance that
the ice or pile moves for each loading cycle is termed as the

Fig. 16. Brittle ice failure of ice sheet (Huang et al., 2007).

third of total thickness. See Fig. 15. This phenomenon is

caused by high level of freedom at top and below surface and
be a typical phenomenon of ice crushing (Jordaan, 2001).
Compared with conventional numerical techniques, such as
finite element method where the failure mode is dependent on
the mesh, these images show the advantages of using a
meshfree method. Fig. 17. FFT of ice failure (v ¼ 50 mm/s).

Please cite this article in press as: Liu, M., et al., Peridynamic simulation of brittle-ice crushed by a vertical structure, International Journal of Naval Architecture
and Ocean Engineering (2016), http://dx.doi.org/10.1016/j.ijnaoe.2016.10.003
 + MODEL

M. Liu et al. / International Journal of Naval Architecture and Ocean Engineering xx (2016) 1e10 9

Fig. 18. FFT of ice failure (v ¼ 130 mm/s). Fig. 21. FFT of ice failure (v ¼ 750 mm/s).

Fig. 19. FFT ice failure (v ¼ 210 mm/s). Fig. 22. FFT of ice failure (v ¼ 1000 mm/s).

Fig. 23. Schematic diagram of characteristic ice failure.

Fig. 20. FFT of ice failure (v ¼ 500 mm/s).

where v is the pile velocity and f is the frequency of the re-
petitive force.
“damage zone length”, see Fig. 23. The average length d for Fig. 24 shows a plot of d=h (h is the level ice thickness)
each loading cycle can be expressed as versus moving pile velocity. The damage zone length for all
simulations maintained between 0.15 and 0.2 times as the ice
v
d¼ ð21Þ thickness. The slightly increasing damage zone length for
f increasing velocities agrees well with the experimental result,

Please cite this article in press as: Liu, M., et al., Peridynamic simulation of brittle-ice crushed by a vertical structure, International Journal of Naval Architecture
and Ocean Engineering (2016), http://dx.doi.org/10.1016/j.ijnaoe.2016.10.003
 + MODEL

10 M. Liu et al. / International Journal of Naval Architecture and Ocean Engineering xx (2016) 1e10

the moving vertical structure. In addition, the lengths of the

damage zones were calculated for each loading cycle. The
damage zone lengths remain 0.15e0.2 times as the ice
thickness. The lengths of the damage zones increase slightly
with increasing moving velocities, even though the signifi-
cance of this scattered data trend is unknown.

References

Bergan, P.G., Cammaert, G., Skeie, G., et al., 2010. On the potential of
computational methods and numerical simulation in ice mechanics[C]//.
IOP Conf. Ser. Mater. Sci. Eng. 10 (1), 012102. IOP Publishing.
Brbaru, F., Yang, M., Alves, L.F., Silling, S.A., Askari, E., Xu, J., 2009.
Convergence, adaptive refinement, and scaling in 1D peridynamic. Int. J.
Numer. Methods Eng. 77 (6), 852e877.
Fig. 24. Plot of d=h (h is the thickness of level ice) versus moving pile Foster, J.T., Silling, S.A., Chen, W.W., 2010. Viscoplasticity using peridy-
velocities. namics. Int. J. Numer. Methods Eng. 81 (10), 1242e1258.
Huang, Y., Shi, Q., Song, A., 2007. Model test study of the interaction between
ice and a compliant vertical narrow structure. Cold Reg. Sci. Technol. 49
even though the significance of the scattered data trend is (2), 151e160.
unknown. Jones, S.J., 1982. The confined compressive strength of polycrystalline ice.
J. Glaciol. 28, 171e177.
Jordaan, I., 2001. J. Mechanics of iceestructure interaction. Eng. Fract. Mech.
6. Conclusions 68 (17), 1923e1960.
Kamio, Z., Matsushita, H., Strnadel, B., 2003. Statistical analysis of ice
The peridynamic model was applied to simulate the inter- fracture characteristics. Eng. Fract. Mech. 70 (15), 2075e2088.
action of level ice and a cylindrical, vertical, rigid structure at Kuehn, G.A., Schulson, E.M., Jones, D.E., et al., 1993. The compressive
different moving velocities. To implement two-body in- strength of ice cubes of different sizes. J. Offshore Mech. Arct. Eng. 115
(2), 142e148.
teractions in the peridynamic model, an interaction force Madenci, E., Oterkus, E., 2014. Peridynamic Theory and its Applications.
density was applied to prevent the material particles that Springer, New York.
belong to different bodies from sharing the same position. To Schulson, E.M., 2001. Brittle failure of ice. Eng. Fract. Mech. 68 (17),
determine the feasibility and accuracy of applying the peri- 1839e1887.
dynamic method on iceestructure interactions, results from Schulson, E.M., 1997. The brittle failure of ice under compression. J. Phys.
Chem. B 101 (32), 6254e6258.
the peridynamic simulation were compared with small-scale Silling, S.A., 2000. Reformulation of elasticity theory for discontinuities and
experiments conducted by Sodhi and Morris. The ice force long-range forces. J. Mech. Phys. Solids 48 (1), 175e209.
records from the numerical simulation show a typically re- Silling, S.A., Askari, E., 2005. A meshfree method based on the peridynamic
petitive nature and the value of the ice forces obtained by model of solid mechanics. Comput. Struct. 83 (17), 1526e1535.
peridynamic simulation is of the same order as that of the Silling, S.A., Epton, M., Weckner, O., et al., 2007. Peridynamic states and
constitutive modeling. J. Elast. 88 (2), 151e184.
small-scale experiments. Visual results show that during Sodhi, D.S., Morris, C.E., 1984. Ice Forces on Rigid, Vertical, Cylindrical
moving of the structure to crush the ice, damaged ice particles Structures. Cold Regions Research and Engineering Lab, Hanover NH.
accumulate in front of the cylindrical structure, which implies Sodhi, D.S., Morris, C.E., 1986. Characteristic frequency of force variations in
the superiority of the peridynamic method to complex material continuous crushing of sheet ice against rigid cylindrical structures. Cold
damage. In addition, in the path cleared by cylindrical struc- Reg. Sci. Technol. 12 (1), 1e12.
Warren, T.L., Silling, S.A., Askari, A., et al., 2009. A non-ordinary state-based
tures, line-like hot spot can be observed, which is high stress peridynamic method to model solid material deformation and fracture. Int.
area arranging ice thickness and occupying one third of total J. Solids Struct. 46 (5), 1186e1195.
thickness. A study of the characteristics of loading cycle and Tian,, Y., Yan, H., 2013. The dynamic ice loads on conical structures. Ocean
Fast-Fourier-Transforms were used to transform the force re- Eng 59, 37e46.
cords from a time to a frequency domain. The characteristic Yue, Q., Guo, F., K€arn€a, T., 2009. Dynamic ice forces of slender vertical
structures due to ice crushing. Cold Reg. Sci. Technol. 56 (2e3), 77e83.
ice failure frequencies increase with increasing velocities of

Please cite this article in press as: Liu, M., et al., Peridynamic simulation of brittle-ice crushed by a vertical structure, International Journal of Naval Architecture
and Ocean Engineering (2016), http://dx.doi.org/10.1016/j.ijnaoe.2016.10.003