ARTICLE IN PRESS

Ocean Engineering 34 (2007) 21242137

www.elsevier.com/locate/oceaneng

Development of structural modeling techniques for evaluating

HDPE plastic net pens used in marine aquaculture
David W. Fredrikssona,, Judson C. DeCewb, Igor Tsukrovc
a

Department of Naval Architecture and Ocean Engineering, United States Naval Academy, 590 Holloway Road, Annapolis, MD 21402, USA
b
Center for Ocean Engineering, University of New Hampshire, 24 Colovos Road, Durham, NH 03824, USA
c
Department of Mechanical Engineering, Kingsbury Hall, University of New Hampshire, Durham, NH 03824, USA
Received 6 February 2007; accepted 18 April 2007
Available online 3 May 2007

Abstract
Finite-element modeling (FEM) techniques are developed to determine the structural capabilities of net pen otation structures
made of high-density polyethylene (HDPE). The modeling approach uses shell elements and localized failure criteria to predict
critical loading conditions. Finite element modeling simulations were performed using values for the modulus of elasticity for weathered
HDPE determined from a series of tensile tests. Poissons ratio and yield stress used with the approach were obtained from standard
manufacturers values. To investigate the method, a series of experiments were performed in the laboratory by testing circular sections
of HDPE pipe to localized failure (kinking). The same test was replicated with the FEM using localized failure techniques.
Laboratory and model results were within 16%. A FEM was then built to represent the complex geometry of a net pen otation structure
deployed at an operational sh farm located in Eastport, ME, USA. Simulations were performed using attachment line tension
values measured at the site. The goal was to assess otation pipe stress levels for typical operational conditions. Simulations were
also performed to investigate the maximum capabilities of the net pen structure with different attachment line congurations.
Quantifying the operational limits will become more important as these systems are considered for more exposed, energetic
environments.
r 2007 Elsevier Ltd. All rights reserved.
Keywords: Coastal aquaculture; Aquacultural engineering; Finite-element analysis; Fish cages

1. Introduction
Most of the existing marine aquaculture installations are
presently located in protected near-shore areas. Many in
the industry are considering moving operations into
exposed sites because of coastal multi-use issues. The
development of inshore marine aquaculture technology has
evolved over the past several years (decades) primarily
through operational experience. It is now common to nd
otillas of surface gravity cages constructed using highdensity polyethylene (HDPE) pipe in most marine nshproducing countries. The use of HDPE pipe for this
application is advantageous because the components are
Corresponding author. Tel.: +1 410 293 6434.

E-mail addresses: fredriks@usna.edu (D.W. Fredriksson),

jcdc@unh.edu (J.C. DeCew), igor.tsukrov@unh.edu (I. Tsukrov).
0029-8018/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.oceaneng.2007.04.007

relatively easy to fabricate and repair, the plastic has

outstanding corrosion characteristics and the material is
compliant and (historically) inexpensive. As the industry
investigates the option of moving to exposed sites, it is
unclear how the existing equipment will be able to
withstand higher energy environments.
The structural integrity of a surface gravity cage system
is investigated as part of a comprehensive study working
with a marine aquaculture company facility with 20
deployed net pens. The project included the installation
of current meters and load cells to measure the forcing and
resulting loads in components of the farm (Fredriksson
et al., 2006). The components consist of anchor leg
assemblies, net pen attachment lines and HDPE net pen
structures. As part of the project, a uidstructure
interaction model (Tsukrov et al., 2003, 2005) was used
with the eld measurements to assess mooring gear

ARTICLE IN PRESS
D.W. Fredriksson et al. / Ocean Engineering 34 (2007) 21242137

operational limits (Fredriksson et al., 2007). In this study,

the structural capabilities of the HDPE net pens used to
contain the nsh product are examined.
Fish farm component details used in the comprehensive
study were obtained from an operational facility located in
Eastport, ME, USA (Fig. 1). The Eastport site is located in
Broad Cove near the border of New Brunswick, Canada.
Deployed at the site is an array of net pens congured in a
5  4 near-surface mooring grid system. It is oriented in a
northsouth direction and held up with otation elements
(Fig. 2). The surface portion of the farm has the
approximate dimension of 219 m  274 m. Each of the net
pens has a nominal radius of 15.9 m.
The aquaculture site in Broad Cove is inuenced by
the extreme tides of the Bay of Fundy (Dudley et al.,
2000). The predominant forcing on the farm components is
a result of the tidal currents. The tidal currents interact
with the nets and HDPE pipe components to create
drag forces. The loads on each net pen are transferred
to the mooring grid system, ground tackle and anchors
using sets of attachment lines (y-lines). To quantify the
forcing at the site, a current meter was installed at an
exterior farm location (see the focus area in Fig. 2). In
addition to the current meter, four load cells were installed
on the West (W), Southwest (SW), East (E) and Southeast
(SE) y-lines. Instrument positions and y-line components are shown superimposed on the aerial photograph
in Fig. 2.

2125

Each of the 20 net pens at the site is circular and is

constructed with HDPE pipe. The surface otation rim
structure of each net pen consists of two concentric rings of
pipe with a nominal diameter and thickness of 323 and
19.8 mm, respectively. The pipes are held together by
passing them through 48 brackets constructed of two larger
pipe sections (diameter 406 mm, thickness 6.9 mm).
The brackets also incorporate a handrail support assembly
made with 114 mm diameter pipe with a thickness of
5.7 mm with vertical and angled members through which a
circular handrail is passed. These components are shown
in Fig. 3 (a load cell recorder strapped to one of the net pen
stanchions is also shown in the gure).
For most deployments at this site, the net pen
components have been specied appropriately through
years of operational experience. However, the actual
capabilities of these components are rarely quantied.
If the same structures are to be deployed in a more
energetic environment, the point of failure will need to be
predicted to prevent escapement, which can have
dire economic and environmental consequences. Therefore,
the objective is to develop a technique for predicting the
structural integrity of the otation structure of HDPE
net pens. The technique is developed utilizing a combination of laboratory and computer model experiments. Using
the technique, nite-element modeling (FEM) simulations
are performed for loading conditions at the Broad Cove
site. The same model is then used to investigate maximum

Fig. 1. The Broad Cove sh farm is located in Eastport, ME, USA.

ARTICLE IN PRESS
2126

D.W. Fredriksson et al. / Ocean Engineering 34 (2007) 21242137

Fig. 2. The sh farm consists of 20 net pens each with an approximate radius of 15.9 m. Instrumentation was deployed in the SW portion of the farm (see
focus area). Load cells were installed on the W, SW, E and SE y-lines on the southern portion of the pen and a current meter at an external farm location.

loading conditions at estimated yield values for the

material.
The rst step in the approach was to carry out a series of
laboratory tensile tests to determine the modulus of
elasticity for samples of weathered HDPE using a range
of loading rates. The next step was to develop the modeling
procedure for HDPE pipe, including stressstrain characteristics and FEM techniques. This was done by utilizing
shell elements and employing localized failure criteria.
Circular ring experiments were then performed where ve
samples of actual pipe were tested to localized failure in the
laboratory. The laboratory tests were replicated using
FEM simulations. The results were then compared.
Finally, the FEM approach was applied to the more
complex geometry of the pen otation structure deployed
at the Broad Cove site. Calculations were performed to
examine the stresses in the pipe during operation at the
farm and to investigate the maximum load the net pen pipe
assemblies can withstand.

2. Finite-element procedure and material modeling

2.1. Experimental study investigating mechanical properties
of HDPE
Unlike linear elastic materials such as steel, HDPE is
considered viscoelastic because the material behaves
differently depending upon the load rate. A slowly applied
load will cause the HDPE to undergo large deformations as
the polymer bonds of HDPE settle and unwind.
However, if the same load is applied quickly, the bonds do
not have sufcient time to react, resulting in less deformation. As a result, the modulus of elasticity relating stress
and strain can be substantially different. Net pens that are
located in exposed marine environments where waves and
currents are prevalent may have a range of loading rates
that are structure and site specic. To investigate the
modulus of elasticity characteristics for weathered HDPE,
a series of tensile tests were performed using samples of

ARTICLE IN PRESS
D.W. Fredriksson et al. / Ocean Engineering 34 (2007) 21242137

2127

Fig. 3. Components of the net pen rim with a load cell recorder attached to one of the stanchions.

Fig. 4. Samples of HDPE were placed into an Instrons tensile testing device (length dimensions are given in cm).

pipe weathered for several years to various seasonal

elements. Since the actual loading rates at the sh farm
site were not determined, three standard loading rates of
0.0127, 0.127 and 1.27 mm/s (typical of tensile testing
procedures) were used in the experimental tests.
Ten pieces of the weathered HDPE were machined into
the typical dog bone shape having an initial (total)
length of 12.7 cm and a test length of 5.08 cm (Fig. 4).

Three sets of stressstrain tests (each with three replicates)

were conducted at the three loading rates in an Instrons
Schenick Testing System (Model 1350). Also shown in
Fig. 4 is the machine with a specimen prior to testing. The
stressstrain results for the tests are shown in Fig. 5ac for
the 0.0127, 0.127 and 1.27 mm/s loading rates, respectively.
For each set of loading rate tests, a modulus of elasticity
was determined by linearizing the results. The values are

ARTICLE IN PRESS
D.W. Fredriksson et al. / Ocean Engineering 34 (2007) 21242137

2128

Fig. 5. Stressstrain data results (with three replicates) when loaded at (a) 0.0127 mm/s, (b) 0.127 mm/s and (c) 1.27 mm/s.

Table 1
Modulus of elasticity values obtained from tensile using weathered HDPE
samples

2.2. Finite-element analysis of HDPE pipes and localized

failure criteria

Sample

FEM of the HDPE pipe was performed by utilizing shell

elements developed to analyze thin structural components
with high bending stiffness. These elements combine both
membrane and bending resistance to applied loads. It is
assumed that the exural deformation of the shell occurs
according to the KirchhoffLove hypotheses, so that
bending stresses and strains are proportional to the
distance from the mid-surface of the shell. The bilinear
four-node thin-shell elements used in the analysis have six
degrees of freedom per node. Bilinear approximation of the
displacements provides continuity of the displacement
functions through the boundary of the element (see for
example, Zienkiewicz et al., 2005). Models were developed
in the commercially available nite-element program
MSC.MARC (www.mscsoftware.com). Models constructed in the software program were denominated as
type 139 as dened in the MSC.MARC element library and
follow a coordinate system as shown in each resulting
gure. Results from the simulations are presented in the
form of von Mises stress for comparison with yield values.
von Mises stress is an equivalent scalar representation of
the calculated stress tensor that characterizes the strain

Modulus of elasticity (Pa)

Rate 1, 0.0127 mm/s

Rate 2, 0.127 mm/s

Rate 3, 1.27 mm/s

1
2
3

7.80  108
5.10  108
7.13  108

9.52  108
6.73  108
9.27  108

1.18  109
7.72  108
1.17  109

S.D.
average

1.41  108
6.67  108

1.54  108
8.37  108

2.33  108
1.04  109

provided in Table 1. Also provided in Table 1 are the

standard deviation and average of the modulus of elasticity
values. Tensile testing results show a dependence on the
loading rate, though the results from using specimen 2 were
lower than those of specimens 1 and 3 for each of the test
sets. It is possible that the difference can be attributed to
varied weathering conditions. The average value of the
modulus of elasticity for the slowest loading rate, equal to
6.67  108 Pa, was used as input for the FEM calculations.
It should be noted that a value 8.00  108 Pa was used by
Vikestad and Lien (2005) to examine the bending stiffness
of sh cage collars.

ARTICLE IN PRESS
D.W. Fredriksson et al. / Ocean Engineering 34 (2007) 21242137

2129

energy of distortion per unit volume. It has been shown to

adequately predict plastic ow and failure of structural
materials.
To model the HDPE pipes used as net pen structural
members, a simplied material model was used. The
modulus of elasticity of the material was assumed to be
constant as long as the material is within its elastic region
(i.e. stress is not larger than the material yield limit). The
yield strain of a typical ductile polymer ranges from 5% to
10% strain (where a metal yields less than 0.1% strain).
Once the stress in the material passes the yield stress,
plastic deformation begins to occur (where the material will
no longer return to its original shape after the load is
removed). The material then follows a stressstrain
relationship similar to that shown by the dashed curve in
Fig. 6. The material will eventually fail or fracture as the
strain increases. Modeling the stressstrain characteristics
is possible but requires an extensive number of tests to
resolve the plasticity of the material under various loading
rates. Instead, the simplied structural model, shown as the
solid curve in Fig. 6, was chosen to represent the HDPE
behavior. In this case, the critical point to determine failure
is the yield stress s(y).
Thus, nite-element analysis of the HDPE pipe requires
the following material parameters: modulus of elasticity
(E), Poissons ratio and yield stress. The modulus of
elasticity was determined from the series of tensile tests
using weathered HDPE samples as described in Section 2.1.
Values for Poissons ratio were taken from manufacturers
standards as 0.42. A sensitivity study was performed
changing Poissons ratio value in a series of FEM
simulations and the results compared. It was found that
the results did not vary signicantly at the time steps used
in the model. This result was similar to that found by
Suhey et al. (2005) in their numerical model study of
membrane structures used for open ocean aquaculture
cages. Yield stress was taken from the standard manufacturer published value of 2.413  107 Pa. In this study,
weathered HDPE components are considered, but according to Ollick and Al-Amir (2003), weather does not
signicantly affect the yield stress of HDPE.
In the FEM procedure, localized failure criteria are
applied as follows. Loads acting on a set of closely spaced

shell elements are slowly increased. When the yield stress at

a certain element is reached, the model distributes the load
to the surrounding elements (effectively decreasing the
stress at the time step). The load is increased and the
procedure is repeated until enough elements, consisting of
approximately half the circumference of the pipe, reach the
yield stress. At this point, it is assumed that the crosssection of the pipe fails, and the simulation is terminated.
Using this simplied representation will have the effect of
over-predicting the material failure (conservative approach) because the polymer will be ductile at this point,
but may still have limited load-bearing capabilities,
whereas the model will not. The predictions of critical
stresses, using this criterion, are dependent upon the niteelement mesh. In the analyses, multiple simulations were
conducted, each time decreasing the size of the elements,
until a negligible stress difference occurred.

Fig. 6. General characteristics of a viscoelastic material (dashed) and the

simplied model assumed for the numerical procedure (solid).

Fig. 7. (a) Test setup for the weathered circular ring experiments. (b) Each
sample was pulled until localized failure occurred.

2.3. Circular ring experiments

To investigate the effectiveness of the FEM localized
failure criteria and the applicability of the tensile tests, a set
of experiments were conducted using circular rings of
HDPE pipe. Tests were rst conducted in the laboratory
where circular rings of HDPE pipe were deected until the
yield point. Then the laboratory tests were simulated using
the FEM technique.

ARTICLE IN PRESS
D.W. Fredriksson et al. / Ocean Engineering 34 (2007) 21242137

2130

The laboratory experiment consisted of ve individual

pull tests performed on 3.66 m diameter circular rings of
HDPE pipe (pipe diameter 88.9 mm; pipe thickness 5.74 mm) having the same exposure as the material
used in the tensile tests. The rings were xed at two

Fig. 8. Results of the tests showed that localized failure occurred at an

average point of 1557 N (excluding results from specimen 1).

Table 2
Geometric and material properties of HDPE pipe used in the circular ring
FEM calculations
Parameter

Value

Overall diameter
Pipe diameter
Wall thickness

3.66 m
8.89 cm
0.57 cm

Material
Modulus of elasticity
Poissons ratio
Yield stress

HDPE
6.67  108 Pa
0.42
2.41  107 Pa

locations on one side spaced approximately 2.8 m (along

the arc). On the other side, a single rope with an inline load
cell was attached to the ring. A photograph of the setup is
shown in Fig. 7a. A come-a-long was slowly used until
local failure (kinking) was observed (Fig. 7b). It was
difcult to quantify the loading rate with this setup since
the experiment was performed manually, however, ve
replicate tests were conducted. The results are shown in
Fig. 8 where the measured tension is plotted as a function
of diametrical deection. Tests using specimens 25 showed
similar results with localized failure occurring at a load of
approximately 1557 N.
The next step was to perform numerical simulations of
the circular ring laboratory tests using the MSC.MARC
FEM software and compare results. The FEM was rst
constructed using the geometric properties of the pipe from
the circular ring laboratory tests. The material properties
used in the model consisted of the published values for
yield stress of 2.413  107 Pa, a Poissons ratio of 0.42 and a
modulus of elasticity equal to 6.67  108 Pa obtained from
the tensile tests. Parameters and values used in the FEM
simulation are provided in Table 2. A mesh of the pipe was
created in the software program using 2540 nodes with
2540 rectangular shell elements. The shell elements in the
vicinity of the load application points were given the
dimensions of 6.05 mm  12.7 mm. The elements at locations away from the load application points were given the
dimensions of 120 mm  12.7 mm. On the mesh, the
loading and boundary conditions were set in the same
location as the laboratory tests.
The model was constrained by xing one line location
in the x, y and z directions. However, to replicate the
degrees of freedom of the ring in the laboratory tests, the
other line attachment location was only xed in the
direction of the point load (Fig. 9). This allowed the model
to bend and ex due to the application of the load and not
restrict any motion. Applied point forces, distributed over
27 nodes, linearly increased as described in Section 2.2. It is
important to note that the load in the FEM simulations

Fig. 9. Using the values in Table 2, FEM calculations were performed to simulate the weathered ring experiments.

ARTICLE IN PRESS
D.W. Fredriksson et al. / Ocean Engineering 34 (2007) 21242137

2131

Fig. 10. A visual comparison between the localized buckling of the weathered ring from the laboratory experiment and the FEM approach.

Fig. 11. A plot showing the load versus the percentage of pipe circumference element failure. The gure also shows the progression of stress contours using
the localized failure technique.

was applied quasi-statically so that the rate of loading was

accounted for by the choice of material parameters only.
During the simulation, the rim deformed until the yield
stress criteria were met. The original and deformed models
of the rim are shown in Fig. 9. The load at failure was
calculated to be 1877 N, approximately 16% greater than
the laboratory tests. Fig. 10 shows a visual comparison of
the kinking observed from both the laboratory and the
computer model tests.
Although in the study the failure criteria are set
when 1/2 the pipe circumference elements yield, the failure

of a few elements can constitute concern. Fig. 11 shows

the total applied load on the rim versus the percent of
pipe circumference failure. Some residual stress existed
at locations around the rim, but the highest levels
were at load application points. This plot shows the
amount of rim failure as a function of load, at the
load application point. It can be seen that once failure
occurs in 25% of the pipe, total collapse occurs shortly
thereafter. From these results, a certain amount of
condence was obtained to further use this modeling
approach.

ARTICLE IN PRESS
2132

D.W. Fredriksson et al. / Ocean Engineering 34 (2007) 21242137

3. Modeling of the net pen otation structure

3.1. Boundary conditions and geometric and material
properties
A mesh was constructed of the entire net pen otation
structure using the geometric and material properties
provided in Table 3. Shell element dimensions varied

Table 3
Geometric and material properties of the net pen otation structure FEM
Component

Parameter

Value

Net pen

Overall diameter
31.8
Circumference
100
Material
HDPE
6.67  108
Modulus of elasticitya
Poisons ratioa
0.42
Yield stressa
2.41  107 Pa

Flotation pipe (inner and outer) Pipe diameter

Thickness

0.3238 m
0.0198 m

Rim sleeve

Pipe diameter
Thicknessb

0.4064 m
0.0690 m

Sleeve support

Pipe diameter
Thickness

0.3238 m
0.0198 m

Handrail

Pipe diameter
Thickness

0.1143 m
0.0057 m

Handrail sleeve

Pipe diameter
Thicknessb

0.1413 m
0.0242 m

Vertical support

Pipe diameter
Thickness

0.1413 m
0.0134 m

Angled support

Pipe diameter
Thickness

0.1143 m
0.0108 m

a
All components used in the model have the same modulus of elasticity,
Poissons ratio and yield stress.
b
Sleeves take the rim or handrail pipe thickness into account.

depending upon the component. The inner and outer

otation pipe assemblies were modeled with elements
having the dimensions of 294.8 mm  79.3 mm. A total of
24,864 nodes and 25,824 elements were used in the model.
A portion of the FEM of the net pen is shown in Fig. 12.
Boundary conditions for the model were similar to those
used for the circular ring laboratory tests, but the load
application was modied. When deployed in the eld, the
primary loads on a net pen are the drag forces associated
with the net and the counteracting y-line tensions. It is
difcult to model the distributed load of the net (around
the rim) and localized forces due to the y-lines simultaneously. Therefore, two load cases were used to investigate
each scenario. Together, these load cases should bracket
the predicted stresses on the system.
The rst load case (LC #1) examines the system from a
global perspective. This approach utilizes a distributed
load representing the drag forces of the net chamber, with
xed and supported boundary conditions representing the
y-line attachment points (Fig. 13a). Similar to the circular
ring, one attachment point was xed in the x, y and z
directions and the other was restricted in the load
application direction. The linear distance (around the
rim) between points was dependent upon the input loading
scenario (for example, if one or two y-lines were supporting
the system in the eld). The load representing the drag of
the nets was assumed to be distributed evenly over the
inner and outer rims. A linearly increasing force was
applied to over 474 nodes. Note that the weight of the nets
and coinciding buoyant force of the rims was not taken
into consideration when modeling these structures.
The second load case (LC #2) investigated the response
of the system due to y-line tensions that can occur at the
sh farm site. Operational personnel secure each net pen
with a gure eight knot consisting of multiple wraps
around the inner and outer otation pipes and secured with
half-hitches. In the FEM model, the tensions in the
mooring line were represented by point loads, which were

Fig. 12. FEM representation of the surface rings used for the net pen otation structure.

ARTICLE IN PRESS
D.W. Fredriksson et al. / Ocean Engineering 34 (2007) 21242137

applied to 1/2 the pipe circumference and evenly distributed on the inner and outer otation pipe to represent
contact similar to the gure eight knot. Note that the
actual load distribution between the pipes may be
dependent upon the number of wraps. The model also
contained two xed points similar to LC #1 (Fig. 13b).
Modeling the structure in this manner addresses a specic
attachment conguration to examine localized stresses. In
addition, these four sets of point loads can have different
magnitudes, representing more eld-like conditions.
Both load cases, however, do assume the y-line locations
(either as constraints or load application points) are
symmetrically located around the rim. For each load case
scenario, the geometric and material properties were
assumed to be the same (Table 3).

2133

3.2. Broad Cove farm case study

An important part of the study was to relate modeling
results to actual operational conditions at the marine sh
farm located in Eastport, ME, USA. As previously
described, the FEM investigations presented here where
part of a larger study that included an extensive set of eld
measurements were current meters and load cells were
deployed at the site (Fredriksson et al., 2006). The current
direction was an important factor in selecting the load
cases used for comparison with the FEM. As part of the
eld study, a current meter was deployed approximately
23 m to the southwest of the farm. Velocity data sets were
examined and specic ones chosen where the net pen was
set-back against the two attachment y-lines, insuring that

Fig. 13. (a) Load case #1, the drag of the nets was assumed to be evenly distributed over the inner and outer rims; (b) load case #2, tensions in the mooring
lines were represented by point loads.

ARTICLE IN PRESS
D.W. Fredriksson et al. / Ocean Engineering 34 (2007) 21242137

2134

the lines were taking the majority of the load. Two data
sets with these loading conditions occurred between 14:00
and 14:20 UTC (coordinated universal time) on 5-May-04
and 22:00 and 22:20 UTC on 16-May-04 where forcing
velocities were relatively consistent (Table 4).
In addition to the current velocity measurements, load
cells data sets were also acquired during the same cases
from the W, SW, SE and E y-lines (see Fig. 2). The
instruments were programmed to measure net pen attachment loads at a rate of 5 Hz for 20 min each hour. The time
Table 4
Velocity measurements obtained from the current meter located at an
external farm location
Load case date

Time (UTC)

East-going (m/s)

North-going (m/s)

05/05/04

13:40
14:00
14:20

0.070
0.271
0.071

0.559
0.137
0.493

05/16/04

21:40
22:00
22:20

0.201
0.296
0.268

0.004
0.304
0.258

series results for each load case (corresponding to the

velocity data sets) are shown in Fig. 14a and b. The average
and maximum loads are provided in Table 5. The
maximum loads are used as input to the structural model
to calculate net pen rim stresses typical of normal
operation.

3.3. Structural modeling results

Using shell elements, the localized failure criteria, and
the previously described boundary conditions, case studies
were performed to examine loads on the net pen structure
used at the Broad Cove site. Using both load cases (LC #1
and LC #2), the otation pipes were subjected to (1) a
distributed load equal to the resultant sum of the two y-line
load vectors, and (2) individual y-line attachment forces.
To model the worst-case scenario, the maximum load of
each recorded value was utilized as provided in Table 5.
The results from the structural model calculations yielded
von Mises stress values for each load case as provided in
Table 6. The maximum von Mises stress obtained using the

Fig. 14. Load cell data from (a) 5-May-2004; 14:00 UTC and (b) 16-May-2004; 22:00 UTC.

ARTICLE IN PRESS
D.W. Fredriksson et al. / Ocean Engineering 34 (2007) 21242137

structural model was 4.23  106 Pa, which is more than ve

times smaller than the 2.41  107 Pa yield value for HDPE.
This result may seem obvious since during normal
operation, the net pen rims rarely fail. However, the
modeling technique now allows the operational limits to be
quantied.

3.4. Estimated maximum load

The next step was to use the structural modeling
technique to determine an estimated failure load of the
net pen rim under various operational circumstances.
Using shell elements and the localized failure criteria,
simulations were performed for three attachment congurations where the net pen was secured with (a) a single
attachment point, (b) two attachment points and (c) four
attachment points, as shown in Fig. 15.
These congurations represent situations ranging from a
worst-case scenario (a) to normal net pen securement (c).
In each one of these load cases, a linearly increasing

Table 5
Average and maximum load cell results from each of the y-line attachment
locations
SW y-line
(kN)

SE y-line
(kN)

E y-line
(kN)

5/05/04
Average 3.21
14:00 UTC Maximum 8.39

1.50
6.26

2.95
3.10

2.41
4.46

5/16/04
Average 2.56
22:00 UTC Maximum 3.07

1.56
2.58

6.60
7.32

3.69
5.13

Date and
time

W y-line
(kN)

Table 6
von Mises stress calculated using the structural model for each load case
and boundary condition
Date and time

Boundary conditions

von Mises stress (Pa)

5/05/04
14:00 UTC

LC #1
LC #2

2.33  106
4.23  106

5/16/04
22:00 UTC

LC #1
LC #2

2.19  106
3.18  106

2135

distributed load, similar to the rst boundary condition

(Section 3.1), was applied to the net pen rim sections until
the yield criteria were met. The net pen rim was xed at
one, two and four locations, respectively. The results of the
model tests are provided in Table 7.
With one y-line attachment point, the net pen rim failed
at 53 kN. However, it is rare that a net pen is secured by
only one point. Typically, this situation exists if there is a
line failure or the net pen is temporarily secured for various
operational reasons. Fig. 16 shows the results when only
one y-line attachment point was examined. The failure load
increased to 98.6 kN when distributed over one full y-line
(or two attachment points). When two y-lines were used,
the failure load increased dramatically. Yield stresses in
stanchions and other sections of the main rims were visible
at 288 kN. However, the pipe still had some structural
integrity at this point. The rim structure buckled at a
load of over 444 kN. These net pen loading conditions can
occur regularly at a farm site if severe environmental
conditions exist. Fortunately, when examining the failure
load for attachment conguration (b) and the rim loads
and associated stresses, the forcing needed to obtain the
failure load is signicant.
It is important to recall that the boundary conditions
used in the model are located symmetrically around the net
pen rim. In the eld, however, operation personnel do not
secure the pens at set locations (equally distributed along
the rim), rather by eye and approximate distances
between the y-lines. This can have an effect on the
structural integrity of the rim. Therefore, a model was
tested similar to LC #1 and y-line attachment conguration
(c), except that one outside y-line attachment point was
moved 4.5 m along the cage circumference, towards the
undisturbed y-line locations. The results showed that
although still adequate for the site, the failure load
decreased by 31%.

Table 7
Estimated failure loads for the three attachment congurations
Load case conguration

y-line attachment points

Failure load (kN)

(a)
(b)
(c)

1
2
4

53.0
98.6
4444

Fig. 15. Three attachment congurations were modeled where the net pen was secured with (a) one, (b) two and (c) four point locations.

ARTICLE IN PRESS
2136

D.W. Fredriksson et al. / Ocean Engineering 34 (2007) 21242137

Fig. 16. The von Mises stress distribution on the net pen rim when 53 kN is applied as a single point load.

4. Conclusion
The technique to structurally model HDPE pipe used in
sh farm net pens shows promise as an effective tool for
specifying these components. As with many modeling
applications, however, these results must be considered
approximate. For example, the value of 6.67  108 Pa used
for the modulus of elasticity may have been underestimated
and should be further investigated by performing additional tensile tests with a larger number of samples. In
addition, the failure criteria are subjective and the
boundary conditions simplied. The question also remains
as to the appropriate loading rate, which affects the
modulus of elasticity for viscoelastic materials.
Quantifying the load rate in environments with waves
and currents may not be trivial. This will become more
important if these structures are considered for more
exposed or open ocean environments. For example, tidal
currents are often considered to be quasi-static, since
forcing periods are on the order of hours. In this
condition, slower loading rates may be considered. In
conditions where intense waves exist, forcing periods are
on the order of seconds, so a faster load rate may be
appropriate. At many open ocean sites, both waves and
currents provide dominant forcing components, so a
combination may also need to be considered. It is clear
that since the structural integrity of HDPE pipe is related
to the loading rate, future work should be conducted to
optimize the approach.
Nevertheless, this present study shows (through normal
operation) that the HDPE pipe used for this near-shore
sh farming application is appropriate. The structural

modeling techniques developed, however, enables the

evaluation of acceptability margins. In the Broad Cove
case, the estimated operational von Mises stress values
were more than ve times less than the yield stress. For
the maximum load condition, it was necessary to consider
the attachment conguration, where the worst-case
situation would occur as a point load. Utilizing two sets
of y-lines (four attachment locations) minimizes the
chance of this occurring. Using the developed techniques,
additional studies can be conducted to optimize net pen
designs and attachments (other than the gure eight
knot) for specic environments, especially if similar
technologies are considered for more exposed, open ocean
conditions.

Acknowledgments
The authors acknowledge the operational and management staff at the Heritage Salmon site in Broad
Cove located in Eastport, ME, USA. The authors also
acknowledge Professors Todd Gross, Barbaros Celikkol,
Rob Swift and Ken Baldwin of University of New
Hampshire for their advice and support. Additional
thanks go to Jim Irish, Glenn McGillicuddy and
Chad Turmelle for technical assistance (also at the
University of New Hampshire). Discussions with Professors Mayer and Hudson (United States Naval Academy)
were valuable. Funding for this project was obtained
through the National Oceanic and Atmospheric Administration, SaltonstallKennedy program under Grant
NAO3NMF4270183.

ARTICLE IN PRESS
D.W. Fredriksson et al. / Ocean Engineering 34 (2007) 21242137

References
Dudley, R.W., Panchang, V.J., Newell, C.R., 2000. Application of a
comprehensive modeling strategy for the management of net-pen
aquaculture waste transport. Aquaculture 187 (3), 319340.
Fredriksson, D.W., DeCew, J.C., Irish, J.D., 2006. A eld study to
understand the currents and loads of a near shore sh farm. In:
Proceedings of the Oceans 2006 MTS/IEEE Conference, Boston, MA,
September 19, 2006.
Fredriksson, D.W., DeCew, J.C., Tsukrov, I., Swift, M.R., Irish, J.D.,
2007. Development of large sh farm numerical modeling techniques
with in-situ mooring tension comparisons. Aquacultural Engineering
36, 137148.
Ollick, A.M., Al-Amir, A.M., 2003. Weathering effects on mechanical
properties of low and high density polyethylene pipes used in irrigation
networks. Alexandria Engineering Journal 42 (6), 659667.

2137

Suhey, J.D., Kim, N.H., Niezrecki, C., 2005. Numerical modeling and
design of inatable structuresapplication to open ocean aquaculture
cages. Aquacultural Engineering 33 (4), 285303.
Tsukrov, I., Eroshkin, O., Fredriksson, D.W., Swift, M.R., Celikkol, B.,
2003. Finite element modeling of net panels using consistent net
element. Ocean Engineering 30, 251270.
Tsukrov, I., Eroshkin, O., Paul, W., Celikkol, B., 2005. Numerical
modeling of nonlinear elastic components of mooring systems. Special
issue on open ocean aquaculture engineering. IEEE Journal of Oceanic
Engineering 30 (1), 3746.
Vikestad, K., Lien, E., 2005. Bending stiffness of a clamps-connected
2-rings PE sh cage collar. In: Proceedings of the 12th International
Congress of the International Maritime Association of the Mediterranean (IMAM 2005), vol. 2, Lisboa, Portugal, pp. 12931299.
Zienkiewicz, O.C., Taylor, R.L., Zhu, J.Z., 2005. The Finite Element
Method: Its Basis and Fundamentals. Elsevier Butterworth-Heinemann.