Composites Part C: Open Access 2 (2020) 100045

Contents lists available at ScienceDirect

Composites Part C: Open Access

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

Design and development of a filament wound composite overwrapped

pressure vessel
Shah Alam a,∗, Gregory R. Yandek b, Richard C. Lee b, Joseph M. Mabry b
a
Department of Mechanical Engineering, Texas A&M University-Kingsville, TX 78363, United States
b
Air Force Research Laboratory, Rocket Propulsion Division, Edwards AFB, CA 93524, United States

a r t i c l e i n f o a b s t r a c t

Keywords: A detailed design optimization of a composite over wrapped pressure vessel (COPV) is performed through finite
Composite overwrapped pressure vessel element analysis (FEA) and testing. The design optimization of a type IV polymer-lined composite overwrapped
Finite element method pressure vessel is performed by means of the finite element method using ABAQUS software. A plug-in named
Burst pressure
WCM in ABAQUS is used that automates the entire process of constructing a COPV FEA model. This work inves-
Failure criteria
tigates the influence of winding angle, number of layers, and sequence of layers on the burst strength of COPV.
Maximum stress, maximum strain, Tsai Hill, Tsai Wu and Hashin progressive failure criteria are used to find the
burst strength of the COPV. Comparing all the results obtained using above the failure criteria, it is found that
the maximum strain criterion is more conservative, and the Hashin criterion is the least conservative. Digital
Image Correlation (DIC) was utilized during burst testing to quantify the failure strain in different directions. The
numerical results were compared with the experimental burst testing results and a reasonably good comparison
was found.

Introduction dious and very time consuming. Therefore, FEA process is used to avoid
multiple iterations.
Composite overwrapped pressure vessels are currently used by Pressure vessels are classified as follows: fully metallic tank (Type
aerospace organizations, for containing high-pressure fluid in the I), metallic liner hoop-wrapped with composite (Type II), metallic liner
propulsion system, medical research centers for life support applica- fully overwrapped with composite (Type III), polymer-lined composite
tions, and science investigations [1]. All COPVs provide great benefit vessel (Type IV), and linerless composite pressure vessel (Type V) [3].
over all metal pressure vessels in terms of weight, but COPVs require The liner of a type IV COPV cannot share part of the load to which the
unique and proper mechanical design, fabrication, and testing require- COPV is subjected. It is used for a non-permeable barrier to prevent
ments. The most noteworthy attribute of COPV is that they require a leakage. In this paper, type IV COPVs are addressed.
more intricate mechanical consideration because of the interchange be- An investigation into non-load sharing liners for COPVs was con-
tween the composite layers and the inward liner. A metallic liner is or- ducted to study the burst failure of the filament wound pressure vessel
dinarily utilized as a part of a COPV as a barrier to fluid penetration composed of T-800 graphite/epoxy [4]. They predict the propagation of
which requires conventional metallic pressure vessel design. Different damage after initial failure and the burst pressure of the filament wound
failure criteria exist for COPVs, and consideration of the factors of these pressure vessel as a function of applied load. In order to predict the burst
criteria are much more different than when they are used for metallic pressure, a progressive failure analysis was performed by using a failure
pressure vessels. criterion with a property degradation model.
Despite their wide-spread use, the development of many COPV de- A three-dimensional elastic analysis of a multi-layered thick-walled
signs is still rather primitive. Often only limited analysis is performed filament-wound hydrogen storage vessel was performed in [5]. The ex-
to obtain an initial design, and then the design is refined through sev- act solution to the stresses on the liner was calculated based on the
eral “build and burst” iterations [2]. However, the cost in material and Lekhnitskii’s theory and plain strain assumptions. An aluminum liner
resources to fabricate multiple test articles is extremely high. To reduce was used to carry out the analysis and for various thicknesses of the
costs, Finite Element Analysis (FEA) is often employed to reduce the liner, shear stress and twisting rate in a normal plane were calculated.
number of iterations required. If done manually, this process can be te- The results were compared with those from previous researchers which

∗
Corresponding author.
E-mail address: [email protected] (S. Alam).

https://doi.org/10.1016/j.jcomc.2020.100045
Received 19 June 2020; Received in revised form 2 September 2020; Accepted 11 September 2020
2666-6820/© 2020 The Author(s). Published 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/)
S. Alam, G.R. Yandek, R.C. Lee et al. Composites Part C: Open Access 2 (2020) 100045

Fig. 1. Geometry of the ASTM D 2585 COPV.

showed that a liner reduces the magnitude of stress development during The effect of impact on the cylindrical part of a pressure vessel is
pressurization. studied in [10]. The study quantified the residual burst pressure of com-
Stress and damage analysis of COPVs with aluminum liners was car- posite cylinders after mechanical impact. More than 200 impact test re-
ried out using numerical analysis subjected to internal pressure [6]. Four sults were assimilated to perform the analysis. The analysis addresses
failure modes for the composite material were considered: fiber tension, the cylindrical characteristics including volume, pressure, liner mate-
fiber compression, matrix tension, and matrix compression. The effect rial, geometry, manufacturer, and testing conditions such as pressure,
of damage evolution to determine the burst pressure and auto-frettage impact energy, etc.
pressure are analyzed in the paper. They concluded that matrix crack A 3D parametric finite element model was analyzed to predict the
damage appears on the head of the vessel while fiber tensile damage damage evolution and failure strength of composite hydrogen storage
appeared on the middle part of the cylinder when subjected to the ser- vessels subjected to internal pressure [11]. Only the cylindrical part of
vice pressure. Auto-frettage exerts a small effect on stress amplitude but the pressure vessel was considered for FEA. They used 6061-T6 alu-
a significant effect on the mean stress. minum alloy as liner and 10 layers carbon fiber/epoxy composite mate-
Research was performed to examine the effects from the storage and rial for the cylinder. Four failure criteria were used to predict damage.
transport of liquefied petroleum gas (LPG) within composite cylinders The analysis concluded that Tsai-Wu failure criteria is the most accurate.
in comparison with those fabricated from steel [7]. The analysis was Liner behavior during pressurization and induced failure was investi-
performed on a composite cylinder fabricated from cross plies. They gated by numerical simulation and testing [12]. The liner used in this re-
compared the results by applying loads on both a steel and a composite search was a polymer blend of 95 wt.% low linear density polyethylene
cylinder model in a finite element software package. Internal pressure, (LLDPE) and 5 wt.% of high-density polyethylene (HDPE). The pressure
axial load, radial load, and impact load were applied to both models. vessel was manufactured by filament winding with variable thickness
FEA predicted that the composite pressure vessel has a higher stress carbon/epoxy composite.
carrying capacity. A study of the cylindrical section of carbon fiber reinforced polymer
A non-linear finite element model was utilized to analyze a stor- composite pressure vessels having four layers was reported in [13]. Both
age pressure vessel subjected to high pressure and local flame impinge- the hoop and helical winding of the fiber was considered. They predicted
ments [8]. The model system includes various sub-models to predict the the burst pressure as a function of various fiber orientation angles. They
temperature evolution of the composite cylinder wall. The analysis of concluded that a ± 45° orientation angle is optimum among all fiber
stresses, failure type and temperature profile were reported. orientation angles analyzed.
An analytical solution and FEA were used to optimize pressure vessel Another study considered pressure vessels with thermoplastic liner
design parameters [9]. Stainless steel and fiber reinforced composite ma- and glass filament wound reinforcements. The commercial software
terials were used for the comparison. The analysis showed that pressure ABAQUS was used to examine the mechanical behavior of pressure ves-
vessel thickness depends on fiber orientation and material type for a cer- sels under internal pressure to failure. The results obtained from the FEA
tain lay-up configuration. The composite pressure vessel demonstrates were then compared with experimental results [14].
higher weight savings compared to stainless steel pressure vessels. Also, A composite pressure vessel for large scale market applications was
the finite element solution agreed well with the analytical solution. studied in [15]. The vessel consisted of a thermoplastic liner wrapped
S. Alam, G.R. Yandek, R.C. Lee et al. Composites Part C: Open Access 2 (2020) 100045

with a filament wound glass fiber reinforced polymer matrix structure. A

high-density polyethylene (HDPE) liner was selected, and a thermoset-
ting resin was used as matrices in the glass reinforced filament wound
laminate. The Tsai-Wu and Von Mises criteria were used to predict com-
posite laminate and thermoplastic liner failures, respectively. A compar-
ison between FEM simulations and experimental results are discussed in
the paper.
From the above literature review it is found that the low burst pres-
sure, low bending stiffness and low impact resistance are the acute prob-
lems in COPV. In this research, it was tried to find the optimal design
to improve the burst strength of COPV. In this context, this work inves-
tigates the influence of carbon fiber (T800S), winding angle, number of
layers, and sequence of layers on the burst strength of type IV COPV.
Maximum stress, maximum strain, Tsai Hill, Tsai Wu and Hashin dam-
age failure criteria are used to find the burst strength of the COPV. Digi-
tal image correlation technique was used to quantify the failure strain of
the fibers. The predicted results are compared with the testing results. Fig. 3. Geometry of the ASTM D 2585 manufactured COPV.

Numerical model development

Table 1
Mechanical properties of carbon fiber composite
A plug-in named WCM in ABAQUS [16] is used that automates the
(T800/epoxy) [17].
entire process of constructing a COPV FEA model. The WCM code is inte-
grated within the ABAQUS/CAE pre-processor as a plug-in module. The Parameter Notation Value
tank manager dialog box is the main graphical user interface (GUI) for Fiber volume fraction Vf 0.60
the WCM. The tank manager provides options to create multiple domes Young’s modulus in direction 1 E1 (ksi) 25,644
and assign them to a complete vessel. This option allows modeling of Young’s modulus in direction 2 E2 (ksi) 1499
COPVs with different polar openings on each end of the vessel. To create Poisson’s ratio in direction 12 v12 0.33
Shear modulus in direction 12 G12 (ksi) 710
a dome in the tank manager, several parameters must be defined. These
Tensile strength in direction 1 XT (ksi) 488
parameters are collected in the wound composite dome window that is Compressive strength in direction 1 XC (ksi) 250
accessed by either the “create” or “edit” buttons in the tank manager Tensile strength in direction 2 YT (ksi) 14
window. In this work, ABAQUS software with WCM is used for creating Tensile strength in direction 2 YC (ksi) 42
pressure vessel geometry and the composite overwrap. Shear strength in direction 12 SL (ksi) 14

Material model and failure criteria

The mechanical properties used in the finite element model are given damage is assumed to be dominated by the fiber failure. In addition
in Table 1. to the damage initiation criterion, an energy-based damage evolution
To calculate the burst pressure of COPV using ABAQUS, failure cri- criterion was used to characterize the progressive damage of the mate-
teria must be defined. The failure criteria for this analysis are maximum rial. Once the critical energy criterion is satisfied by an element, it is
stress, maximum strain, Tsai Hill, and Tsai Wu [18] and Hashin [19]. removed from the simulation. The internal pressure is applied gradu-
Since those criteria are well known, the details description of the crite- ally with small load increments and the failure index is observed. When
ria is not given here. These five failure criteria are used to compare the the failure index is more than 1, then the applied load is considered the
results. For a COPV under pressure load, the failure mode for Hashin pressure at which the COPV fails.

Fig. 2. Geometry of the ASTM D 2585 Liner.

S. Alam, G.R. Yandek, R.C. Lee et al. Composites Part C: Open Access 2 (2020) 100045

Fig. 4. Half tank quarter finite element model of (ASTM D 2585) and composite layups.

Fig. 5. Maximum strain failure index (MSTRN) less than 1 at load step 0.895.

Following Hashin’s parameters are used in this analysis: the WCM module. The elastic properties of the composite layers used in
this analysis are presented in Table 1. Static structural modeling was
Failure initiation stress [17], XT =488 ksi, XC =250 ksi, YT =14 ksi, carried out based on the finite element method (FEM) with ABAQUS
YC =42 ksi, SL =ST =14 ksi. software. Nonlinear geometry was considered in all cases since large
Longitudinal fracture energy [20], Gtl =142.83 k-lbf in-2 , Gcl = 40.32 and unbalanced deformations are expected. The geometry of the ASTM
k-lbf in-2 D 2585 COPV geometry, liner, and a manufactured COPV are shown in
Fig. 1, Fig. 2, and Fig. 3, respectively.
Transverse fracture energy [20], Gtt =1.019 k-lbf in-2 , Gct =17.07 k-
lbf in-2. Viscosity coefficient [20] 0.25 psi.s for fiber and 0.75 psi.s for
Wrapping layers/ layers of COPV
transverse directions. Where XT , XC , YT , YC , SL , ST are the longitudi-
nal tensile and compressive strengths, the transverse tensile and com-
Every polar winding layer consists of 2 plies while every hoop layer
pressive strengths, and the longitudinal and transverse shear strengths,
consists 1 ply. The COPV lay-up consists of 5 basic plies. (Fig. 4). These
respectively.
are:

Finite element model (i) Ply 1: 0.23-inch bandwidth, 0.033-inch thickness and −13° polar
winding.
Composite overwrapped pressure vessel models were created with- (ii) Ply 2: 0.23-inch bandwidth, 0.033-inch thickness and 13° polar
out liner and fully overwrapped by a T800S carbon/epoxy prepreg using winding.
S. Alam, G.R. Yandek, R.C. Lee et al. Composites Part C: Open Access 2 (2020) 100045

Table 2
Summary of FEA results for ASTM D 2585.

Predicted Burst Pressure range Predicted Burst Pressure

from FEA using different failure average burst from Testing (3
Case No Winding angle/ply sequence criteria (psi) pressure (psi) Vessels) (psi)

1 PHP [−13+13+87.5–13+13] 2225–2237.5 2231.25

2 PHP [−13+13+88–13+13] 2237.5–2300 2268.75 2282, 2327,
2391
3 PHP [−13+13+88.5–13+13] 2250–2262.5 2256.25
4 PHP [−13+13+89–13+13] 2250–2275 2262.50
5 PHP [−15+15+88.5–15+15] 2262.5–2287.5 2275.00
6 PHP [−17+17+88.5–17+17] 2287.5–2312.5 2300.00
7 PHP [−19+19+88.5–19+19] 2312.5–2337.5 2325.00
8 PHPH [−13+13+88.5–13+13+88.5] 3847.5–3892.5 3870.00
9 PPHH [−13+13+13–13+88.5 + 88.5] 3847.5–3915 3881.25
10 HHPP [88.5 + 88.5–13+13] 3847.5–3915 3881.25
11 PHHP [−13+13+88.5 + 88.5–13+13] 3847.5–3893 3870.25

P: polar angle, H: hoop angle.

Fig. 6. Maximum strain failure index (MSTRN) greater than 1 at load step
0.90.

Fig. 7. Maximum stress failure index (MSTRS) smaller than 1 at load step 0.89.
S. Alam, G.R. Yandek, R.C. Lee et al. Composites Part C: Open Access 2 (2020) 100045

Fig. 8. Maximum stress failure index (MSTRS) greater than 1 at load step
0.895.

Fig. 9. Tsai Hill failure index (TSAIH) smaller than 1 at load step 0.90. Fig. 10. Tsai Hill failure index (TSAIH) greater than 1 at load step 0.9050.

integration elements. Fig. 4 displays the FE model and composite lay-up

(iii) Ply 3: 0.23-inch bandwidth, 0.009-inch thickness and 88° hoop
of the ASTM D 2585 vessel.
winding.
It is important to use an appropriate mesh size for obtaining reason-
(iv) Ply 4: 0.23-inch bandwidth, 0.033-inch thickness and −13° polar
able results with optimum usage of computer resources. The choice of
winding.
mesh size might have a slight discrepancy in the output results. A fine
(v) Ply 5: 0.23-inch bandwidth, 0.033-inch thickness and 13° polar
mesh might increase the number of element level calculations and in-
winding.
crease the computational cost but increase the accuracy of the results.
It is a common fact that refinement of the mesh yields better results
The polar angle, the hoop angle, total number of layers, and layer for most simulations. Mesh sensitivity was investigated by varying the
sequences are changed through various finite element models and solved element size and the final optimized mesh is used in the analysis. The
to obtain the maximum burst pressure. mesh was generated using 800 elements along longitudinal profile. The
Constant dome shape and outer dimensions were maintained across composite vessel was divided into 5 segments and each segment was
all FE models (see Figs. 1, 2, 3). The geometry and the model were as- divided into 10 elements which results in total 40,000 quadratic hexa-
sumed axisymmetric. A half tank quarter model (90° sweep) is analyzed. hedral elements with 40,851 nodes.
The composite layers were modelled by changing winding angle by 0.5° The local cylindrical coordinate system was created to assign the
increments until the 90° value was reached at the polar boss. The models boundary conditions. Displacements in the radial and circumferential
were developed using S8R, 8-node doubly thick curved shell, reduced directions were fixed at the polar boss region. The symmetric bound-
S. Alam, G.R. Yandek, R.C. Lee et al. Composites Part C: Open Access 2 (2020) 100045

Fig. 11. Tsai Wu failure index (TSAIW) smaller than 1 at load step 0.895.

Fig. 12. Tsai Wu failure index (TSAIW) more than 1 at load step 0.90.

ary conditions are applied in symmetric planes. A uniform pressure of more than 1 are shown in the gray areas. In Fig. 5, the maximum strain
magnitude 2500 psi was applied to the internal face of the model. failure indices more than 1 are found in a small area of the top con-
straint. In other areas, the index contour values are lower than 1. Hence
Results and discussions this step time of load is not considered as burst pressure.
In Fig. 6, the maximum strain failure indices (MSTRN) at time step
Eleven (11) cases (shown in Table 2) based on the polar and hoop 0.9 are shown and the values more than 1 are shown in the gray areas.
angles, and ply sequences are analyzed in this research to find the best The MSTRN values more than 1 are found in the large hoop areas at this
combination for the maximum burst pressure. The maximum internal step time. Hence this step time of load is considered as the burst pres-
pressure imposed in the FE model for load cases 1–6 was 2500 psi and sure. The burst pressure at this step time is (0.90 × 2500 psi) 2250 psi.
for load cases 7–11 was 4000 psi. The load is gradually applied in the The maximum stress failure indices (MSTRS) are shown in Fig. 7 for
model and the failure index is observed. When the failure index is more the step time of 0.89. The maximum stress failure index values are lower
than 1, then the pressure at that load step is considered to have induced than 1 in all the areas of the model. Hence this step time of load is not
failure. The failure indices for maximum strain, maximum stress, Tsai considered as the burst pressure.
Hill, Tsai Wu and Hashin for load case 2 are shown from Fig. 5 to Fig. 15. In Fig. 8, the maximum stress failure indices (MSTRS) at step time
The maximum strain failure indices (MSTRN) are shown in Fig. 5 and 0.9 are shown. The failure indices are more than 1 in the large hoop
6 for the step time of 0.895 and 0.9, respectively. The failure indices areas (gray). Therefore, this step time of load is considered as the burst
S. Alam, G.R. Yandek, R.C. Lee et al. Composites Part C: Open Access 2 (2020) 100045

Fig. 13. Hashin fiber tension failure index (HSNFTCRT) lower than 1 at load step 0.91.

Fig. 14. Hashin fiber tension failure index (HSNFTCRT) more than 1 at load step 0.92.
S. Alam, G.R. Yandek, R.C. Lee et al. Composites Part C: Open Access 2 (2020) 100045

Fig. 15. Hashin fiber compression failure index (HSNFCCRT) lower than 1 at load step 0.92.

Fig. 16. Polar winding and helical (hoop) winding [17].

pressure. The burst pressure at this step time as per maximum stress (gray). This step time of load is considered as the burst pressure. The
failure criterion is (0.90 × 2500 psi) 2250 psi. burst pressure at this step time is (0.90 × 2500 psi) 2250 psi.
The Tsai Hill failure index (TSAIH) distribution is shown in Fig. 9 for In Fig. 13, Hashin fiber tension damage criterion (HSNFTCRT) con-
the step time of 0.9. The TSAIH failure index values are lower than 1 in tour plot at step time 0.91 is shown. The burst pressure is detected when
all the areas of the model. Hence, this step time of load is not considered the Hashin fiber tension damage value (HSNFTCRT) is more than 1. Due
as the burst pressure. to internal pressure, the outer layer of the COPV is tensile stressed. It
In Fig. 10, the Tsai Hill failure indices (TSAIH) at step time 0.905 are is seen in Fig. 13 that the HSNFTCRT values are lower than 1 at step
shown and the values are more than 1 in the large hoop areas (gray). time 0.91. Therefore, the fiber complete damage is not detected in any
This step time of load is considered as the burst pressure. The burst layers of the COPV. The envelope section values shown in Fig. 13 are
pressure at this step time is (0.905 × 2500 psi) 2262.5 psi. the maximum values of all the three layers. Since the maximum value
The Tsai Wu failure index (TSAIW) contour is shown in Fig. 11 for of the HSNFTCRT is 0.96, the fibers are stressed but not completely
the step time 0.895. The TSAIW failure indices are lower than 1 in all failed.
the areas of the model. Hence, this step time of load is not considered Hashin fiber tension damage criterion (HSNFTCRT) contour plot at
as the burst pressure. step time 0.92 is shown in Fig. 14. It is seen in Fig. 14 that the HSNFTCRT
In Fig. 12, the Tsai Wu failure index (TSAIW) contour at step time values are more than 1 at step time 0.92 in the gray areas. Therefore,
0.90 is shown and the values are more than 1 in the large hoop areas the fibers complete damage of the COPV is detected at this step time.
S. Alam, G.R. Yandek, R.C. Lee et al. Composites Part C: Open Access 2 (2020) 100045

Fig. 17. Variation of burst pressure with hoop

angles.

Fig. 18. Variation of burst pressure with polar

angles.

Hence, this step time of load is considered as the burst pressure. The The burst pressure is considered in this analysis when the fiber is
burst pressure at this step time is (0.92 × 2500 psi) 2300 psi. completely damaged in any layers and hence the matrix damage is not
In Fig. 15, Hashin fiber compression damage criterion (HSNFCCRT) analyzed to determine the burst pressure of the COPV.
contour plot is shown. The complete fiber damage is detected when the Similarly, the burst pressure is calculated for all other cases. A sum-
Hashin fiber compression damage values (HSNFCCRT) are more than mary of the results is shown in Table 2.
1. Due to internal pressure, the inner layer of the COPV is compressive Eleven cases are considered to find the effect of polar and hoop an-
stressed. It is seen in Fig. 15 that the HSNFCCRT values are lower than gles, and the sequences of layers on the burst pressure. The polar and
1 at step time 0.92 in all the areas of the model. Therefore, the fiber hoop angles winding are shown in Fig. 16. The helical winding angles
complete damage is not detected in any layers of the COPV. is called hoop angle in this research. In cases 1–4, the hoop angles are
S. Alam, G.R. Yandek, R.C. Lee et al. Composites Part C: Open Access 2 (2020) 100045

Fig. 19. Variation of burst pressure with polar

and hoop layers sequences.

varied from 87.5 to 89° keeping the polar angle (13°) constant. The lim-
ited hoop angle range (87–89°) is considered based on the filament wind
ability. The variation of the burst pressure with the hoop angles is shown
in Fig. 17. It is found that the burst pressure increases with the increases
of hoop angle. This is reasonable as the hoop angle increases, the tan-
gential hoop stress resistance of the COPV increases, and hence the burst
pressure increases.
In cases 3, 5–7, the polar angles are varied from 13 to 19° keep-
ing the hoop angle (88.5°) constant to find the effect of polar angle
on the burst pressure and find the optimum polar angle. The variation
of the burst pressures with the polar angles is shown in Fig. 18. It is
found that the burst pressure increases with the increase of polar an-
gle. Generally, the longitudinal load is taken by the polar layers and
the thickness of polar layers of COPV is inversely proportional to the
cosine of polar angle [17]. As the polar angle increase, the cosine value
decreases and the thickness of COPV increases. Thus the burst pressure
increases.
In cases 8–11, the sequence of polar (P) and hoop (H) layers are
varied keeping the total number of layers (4) constant to find the effect Fig. 21. TESCOM ER5000.

Fig. 20. Winding process of COPV in the filament winding machine. Fig. 22. Burst test chamber with COPV.
S. Alam, G.R. Yandek, R.C. Lee et al. Composites Part C: Open Access 2 (2020) 100045

Fig. 23. Burst test failure region.

Fig. 24. Burst test failure COPVs.

Fig. 25. Burst test data.

of the sequence on the burst pressure and find the best sequence for the Hydrostatic testing for burst pressure
maximum burst pressure. It is found (Fig. 19) that the sequences PPHH
and HHPP have the maximum burst pressure where PHPH and PHHP The test is conducted in an acrylic walled compartment using a test
have minimum burst pressure. fixture supporting the COPV from top. The high-pressure water and ni-
trogen are injected into the COPV using TESCOM ER5000 pressure con-
Experimental testing troller (Fig. 21). The pressure is gradually increased by 20 psi increment
till burst. The test chamber is shown in Fig. 22.
COPV sample preparation
Testing results
The COPVs were manufactured using a filament winding machine.
The steps of preparation are: (i) fabricating mold for mandrel (ii) fab- Burst pressure measurement
ricating mandrel (iii) wrapping mandrel with Teflon tape (iv) program- Several tests were conducted, and the testing results along with pre-
ming filament winding machine using mandrel geometry (v) filament dicted results are given in Table 2. It is found from the comparison of
winding COPV using T800S-10E-24 K/UF3323 towpreg, supplied by the results that the predicted burst pressures are within 5% of testing
TCR Composites, Inc. (vi) Curing at 350°F for 24 hrs (vii) Removing wa- results. The failure zone and failed COPV are shown in Fig. 23 and Fig.
ter soluble mandrel (vi) Install end flange (vii) install strain gage (viii) 24, respectively. From the observation of the failed COPVs it is found
attached COPV with pressure controller tube in burst test chamber (ix) that the COPVs have failed in the hoop areas by local tensile or shear of
install digital image correlation system (DIC) (x) increase pressure grad- the fiber. The FE predicted failure zone was also in the hoop areas. The
ually until COPV failure. A photograph of the filament winding process polar fiber does not exhibit any type of premature failure or distress.
of COPV is shown in Fig. 20. There was bond failure of the composite in the tension zone in the hoop
S. Alam, G.R. Yandek, R.C. Lee et al. Composites Part C: Open Access 2 (2020) 100045

Fig. 26. Data alignment on the surface of COPV.

Fig. 27. Strain in fiber longitudinal direction (hoop).

areas (Fig. 23). Three burst tests are conducted and the burst test data a non-contact measuring technique based on Digital Image Processing.
are shown in Fig. 25. DIC is capable of mapping 3D coordinates and evaluate displacement
and strain maps on the surface of measured samples. DIC uses a stochas-
tic pattern and/or point marker to track the surface of the materials with
Burst strain measurement
subpixel accuracy. This technique is used for a wide range of applica-
The burst strains were measured using a 3D digital image correla-
tions in materials research and component testing to analyze the static
tion system named ARAMIS. ARAMIS digital image correlation (DIC) is
S. Alam, G.R. Yandek, R.C. Lee et al. Composites Part C: Open Access 2 (2020) 100045

and dynamic behaviors of specimens. The full field strain map gener- Acknowledgment
ated by the DIC algorithm is commonly used to replace strain gauges
and other traditional sensors for simulation (FEA) validation. This work is supported by 2017 Air Force Research Lab Summer
The actual surface data was aligned with the fiber direction as shown Faculty Fellowship Program.
in Fig. 26. A 3–2–1 alignment was used to translate the camera coordi-
nates into fiber direction. X, Y and Z represent fiber longitudinal, trans- References
verse and out of plane directions. Major strain was calculated on the
[1] McLaughlan P., Forth S., and Grimes L. Report, composite overwrapped pressure
entire surface. High stress areas of interest were labels and graphs were vessels: a primer. NASA/SP-2011-573 (2011).
plotted on those areas. [2] J. Thesken, P. Murthy, S. Phoenix, et al., A theoretical investigation of composite
The strain in the fiber longitudinal direction is measured in locations overwrapped pressure vessel (COPV) mechanics applied to NASA full scale tests,
Natl. Aeronaut. Space Adm. (2009) Technical report no. NASA/ TM-2009-215684.
of interest (Fig. 27). Just before the strain became non-linear, the highest [3] D. Lin, J. Hsieh, N. Chindakham, et al., Optimal design of the composite laminate
strain observed was 1.63%. The graphs in different colors correspond to hydrogen storage vessel, Int. J. Energy Res. 37 (2013) 761–768.
strains in different locations. [4] D. Son, S. Chang, Evaluation of modelling techniques for a type III hydrogen pressure
vessel (70MPa) made of an aluminum liner and a thick carbon/epoxy composite for
fuel cell vehicles, Int. J. Hydrog. Energy 37 (2012) 2353–2369.
Conclusions [5] B. Shivamurthy, Siddaramaiah, M. Prabhuswamy, Design, fabrication, and testing of
epoxy/glass-reinforced pressure vessel for high-pressure gas storage, J. Reinf. Plast.
Compos. 29 (2010) 2379–2386.
The present research is performed to optimize the design of type
[6] D. Leh, P. Saffre´, P. Francescato, et al., Multi-sequence dome lay-up simulations
IV polymer lined composite overwrapped pressure vessel ASTM D2585 for hydrogen hyper-bar composite pressure vessels, Compos. Appl. Sci. Manuf. 52
COPV. Towpreg carbon fibers (T800S-10E-24 K) and resin TCR UF3323 (2013) 106–117.
are used for this study. It is found that the 6th model (case 6 in Table 2) [7] J. Park, C. Hong, C. Kim, et al., Analysis of filament wound composite structures
considering the change of winding angles through the thickness direction, Compos.
with a combination of three polar-helical-polar winding (PHP [−17+17 Struct. 55 (2002) 63–71.
+ 88.5–17+17) and 9th model (case 9, Table 2) with four polar-polar- [8] Y. Doh, C. Hong, Progressive failure analysis for filament wound pressure vessel, J.
helical-helical winding (PPHH [−13+13+13–13+88.5+88.5) combina- Reinf. Plast. Compos. 14 (1995) 1278–1306.
[9] F. Rueda, J. Otegui, P. Frontini, Numerical tool to model collapse of polymeric liners
tion provide maximum pressure containment capability. The principle in pipelines, Eng. Fail Anal. 20 (2012) 25–34.
findings are as follows: [10] L. Zu, J. Wang, S. Li, Analysis of multilayered thick-walled filament wound hydrogen
storage vessel, Int. J. Hydrog. Energy 36 (2014) 21083–21096.
(i) Tensile failure is observed in the cylindrical (hoop) areas in most [11] Q. Wu, X. Chen, Z. Fan, et al., Stress and damage analysis of composite overwrapped
of the cases. pressure vessel, Procedia Eng. 130 (2015) 32–40.
[12] N. Al-Huniti, O Al-Habahbeh, Composite LPG cylinder as an alternative to steel cylin-
(ii) The burst pressure obtained from maximum stress and strain cri- ders: finite element approach, in: International Conference on Manufacturing and
teria are more conservative (less) and that from the Hashin dam- Material Processing (ICMM 2006), 2006, pp. 363–368.
age criterion is more liberal. [13] J. Hu, J. Chen, S. Sundararaman, et al., Analysis of composite hydrogen storage
cylinder subjected to localised flame impingements, Int. J. Hydrog. Energy 33 (2008)
(iii) The burst pressure is very sensitive to composite ply thickness 2738–2746.
and winding angle. [14] K. Kamal, T. Sayed, A. Elbutch, et al., Analytical and finite element modelling of
(iv) The burst pressure increases with the increase of hoop and polar pressure vessel for seawater reverse osmosis desalination plants, Desalination 397
(2016) 126–139.
angles.
[15] P. Blanc, Burst pressure reduction of various thermoset composite pressure vessel
(v) During burst testing, just before the strain became non-linear, the after impact on cylindrical part, Compos. Struct. 160 (2017) 706–711.
highest strain observed is 1.63%. [16] https://www.3ds.com/products-services/.../wound-composite-modeler-for-abaqus
(accessed 15 June 2020).
[17] S. Peters, Composite Filament Winding, ASM International, 2011.
Declaration of Competing Interest [18] http://help.autodesk.com (accessed 20 June 2020).
[19] Hashin, Failure criteria for unidirectional fiber composites, J. Appl. Mech. 47 (2)
The authors declare that they have no known competing financial (1980) 329–334.
[20] J.T. Rappolt, Analysis of a Carbon Fiber Reinforced Polymer Impact Attenuator for
interests or personal relationships that could have appeared to influence a Formula SAE Vehicle Using Finite Element Analysis MS Thesis, 2015.
the work reported in this paper.