Accepted Manuscript

Micromechanics-based progressive failure analysis of carbon fiber/epoxy composite

vessel under combined internal pressure and thermomechanical loading

Liang Wang, Chuanxiang Zheng, Shuang Wei, Zongxin Wei

PII: S1359-8368(15)00693-9
DOI: 10.1016/j.compositesb.2015.11.018
Reference: JCOMB 3892

To appear in: Composites Part B

Received Date: 25 August 2015

Revised Date: 8 October 2015
Accepted Date: 20 November 2015

Please cite this article as: Wang L, Zheng C, Wei S, Wei Z, Micromechanics-based progressive
failure analysis of carbon fiber/epoxy composite vessel under combined internal pressure and
thermomechanical loading, Composites Part B (2015), doi: 10.1016/j.compositesb.2015.11.018.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Micromechanics-based progressive failure analysis of carbon fiber/epoxy composite vessel

under combined internal pressure and thermomechanical loading

Liang Wang1,2*, Chuanxiang Zheng1, Shuang Wei1, Zongxin Wei1

PT
1
College of Chemical and Biological Engineering, Zhejiang University, Hangzhou 310027, China.
2
Department of Civil and Environmental Engineering and Department of Mechanical Engineering,

RI
Northwestern University, Evanston, IL 60208, USA.

SC
*Corresponding author. E-mail: wangliangtcdri@zju.edu.cn

U
Abstract：A progressive failure analysis algorithm based on micromechanics of failure (MMF) theory
AN
and material property degradation method (MPDM) is developed, wherein the MMF is used to predict the

failure initiation at constituent level and the MPDM is employed to account for the post failure behavior
M

of the damaged materials. The progress of damage is controlled by a linear damage evolution law, which

is based on the fracture energy dissipating during the process. This micromechanics-based approach is
D

implemented by a user-material subroutine (UMAT) in ABAQUS, which is sufficiently general to predict

TE

the ultimate strength and complex failure behaviors of the composite vessel subject to both high pressure

and thermal loading. In addition, the predictions of the model are also compared with those by experiment
EP

and traditional finite element analysis.

Key words: A. Layered structures; B. Strength; B. Thermomechanical; C. Micro-mechanics

C
AC

1. Introduction

Fibre-reinforced polymer composites are finding increasing application in aerospace, marine and many

other industries due to the advantages they provide in performance, structural efficiency and cost. The

carbon fiber/epoxy composite pressure vessel is an important application of composites in the field of

hydrogen storage and transportation, which contributes to safety, economy and high efficiency [1,2]. In

1
 ACCEPTED MANUSCRIPT

general, the composite vessel can be taken as aluminum-carbon fiber/epoxy composite laminated

structures. The laminate layers are stacked by placing the composite layers with different thickness and

different ply orientations, such to achieve high stiffness and strength of the vessel structure. The

design flexibility of the composite vessel enables it to be applicable to various working conditions by

PT
adapting the ply stacking patterns and the vessel geometry parameters [3].

The optimal design of composite vessel as a fundamental research highly depends on the failure

RI
properties and ultimate strength of the composite structure. However, the fast filling of hydrogen leads to

SC
a significant temperature rise within the vessel due to the Joule Thomson effect and the released heat of

gas compression [4,5]. The composite vessels are directly subjected to the cyclic loading of both high

U
pressure and temperature , which contribute to the complicated failure mechanisms of the vessel structure,

such as fiber breakage, matrix cracking and fiber/matrix interface debonding, from the point view of
AN
composite micromechanics [6,7,8]. There is now a need for reliable failure theories and damage evolution

methodologies which will accurately and effectively predict the complex failure mechanisms of the
M

composite vessel structure.

D

Since the failure of composite materials exhibit significant heterogeneity and anisotropy, various
TE

approaches have been proposed to characterise the onset and progression of damage during the past

decades [9-13]. However, most of the failure theories among them are macroscopic or ply-level failure
EP

criterions which need empirical determination of the different failure modes and the predicted results are

not always precise and reliable. By contrast, the micromechanics-based failure theory determine material
C

failure at the constituent level, which show more accuracy in predicting the complicated failure behaviors
AC

especially for the composite vessel subjected to coupled thermal-mechanical loadings. Typical examples

of the micromechanics-based failure theories are the strain invariant failure theory (SIFT) by Gosse et al.

[14] and the multicontinuum theory (MCT) by Garnich and Hansen [15]. Recently, Ha et al. [16]

proposed a micromechanics of failure (MMF) criterion based on comprehensive failure theory of

homogeneous isotropic constituents. Essentially, it is the application of a quadratic-type failure criterion

at the local points of micromechanics-unite cell model, in which the fibers and matrix are modeled

2
 ACCEPTED MANUSCRIPT

explicitly. The micromechanics of failure theory is usually performed with a micromechanical calculation

in order to obtain the interaction information between the micro- and macro-level of composites.

For modeling the post-initial failure behaviors of the composite laminated structure, progressive failure

analysis of composites are mainly performed based on continuum damage mechanics (CDM), in which

PT
the loss of stiffness can be physically considered as a consequence of distributed microcracks and

microvoids and these material defects are phenomenologically represented by introducing a set of internal

RI
state variables [17-21]. Within the framework of CDM, the material property degradation method can be

SC
successfully applied to approach stiffness reduction in the MMF based progressive failure analysis. Since

the MMF approach is developed in homogeneous constituent level, the degradation scheme is much

U
simpler requiring a single degradation parameter for the matrix and fiber respectively [22]. Once the

damage is detected in the fiber or matrix, the corresponding damage variable will degrade to some extent
AN
based on the prescribed damage evolution strategy.

The objective of this study is to propose a micromechanics-based progressive damage analysis

M

strategy, which is sufficiently general to predict the coupled thermal-mechanical responses and complex
D

failure behaviors of the composite vessel structure. An effective finite element model is developed based
TE

on the integration of micromechanics of failure (MMF) theory and material property degradation method

(MPDM), where the MMF is used to predict the failure initiation at each constituent and the MPDM is
EP

employed to account for the post failure behavior of the damaged materials. In addition, the

micromechanics analysis of a typical unit cell model is performed in order to obtain the interaction
C

information which bridge the micro- and macro-level analysis. The predicted results
AC

show good consistency with those of traditional finite element analysis and experiments. This

micromechanics-based progressive failure analysis provides a deeper insight into the multiple failure

mechanism and complex thermal-mechanical behaviors of the composite vessel structure.

3
 ACCEPTED MANUSCRIPT

2. Damage progression model

As the mismatched material properties of fiber and matrix, any external mechanical and thermal

loading will result in a non-uniform micro-stress distribution at the constituent level of composite

materials. The damage initiation and propagation can have different mechanisms depending on where the

PT
failure exists, i.e., in the fiber or in the matrix. In this study, a micromechanics of failure (MMF) criterion

for each constituent is adopted to determine where the failure initiates. The constituent properties of the

RI
failed layer or element are then degraded with the given degradation factors, i.e. dm for the matrix damage

SC
and df for the fiber damage, to model the post-initial failure behavior of the material. This approach for

modeling micromechanics-based progressive failure analysis of composites is implemented by a user-

U
defined material subroutine (UMAT) in ABAQUS [23]. Details of these modelling strategies are

presented in the following sections.

AN
2.1 Micromechanics-baesd failure criterion

The micromechanics of failure (MMF) criterion was developed by Ha et al. [16] for failure prediction
M

of composites at the constituent level. In the MMF criterion, fibre is considered to be longitudinally
D

continuous and has a higher modulus and strength than those of matrix, which suggests that fiber supports
TE

almost the entire load for both longitudinal tension and compression. Thus, fiber dominated damage

initiation is determined using a simple non-interacting Max-stress criterion:

EP

(1) Fibre failure criteria

−C f < σ f 1 < T f , (1)

C

where Tf and Cf are fiber tensile and compressive strengths, and σ f 1 is the fiber micro stress in the
AC

longitudinal direction.

The matrix is assumed as isotropic material, and has different tensile and compressive strengths. A

modified von Mises failure criterion is adopted to determine the initiation of matrix damage:

(2) Matrix failure criteria

4
 ACCEPTED MANUSCRIPT

σ vm
2
1 1
+( − ) I1 = 1, (2)
TmCm Tm Cm

where Tm and Cm are matrix tensile and compressive strengths respectively, I1 and σ vm are the first stress

invariant and von Mises stress at micro level.

PT
Note that, the micro stress caused by the macro mechanical and thermal stress is due to the material and

geometric inhomogeneity of composites. The relationships between macro and micro stress at constituent

RI
level, i.e. in the fiber or matrix, can be described by introducing a set of stress amplification factors, M

SC
and A：

σ = Mσ σ + Aσ ∆T (3)

 M 11

U
M 12 M 13 M 14 0 0 
M 0 
AN
M 22 M 23 M 24 0
 21 
 M 31 M 32 M 33 M 34 0 0 
Mσ =   (4)
 M 41 M 42 M 43 M 44 0 0 
M

 0 0 0 0 M 55 M 56 
 
 0 0 0 0 M 65 M 65 
D

Aσ = [ A1 A6 ]σ
T
A2 A3 A4 A5 (5)
TE

where M and A are mechanical and thermal-mechanical stress amplification factors, respectively.

Generally, these stress amplification factors can be calculated with a unit cell model either analytically or
EP

numerically. Details about the typical unit cell model and the calculation of stress amplification factors for

the given composite structure are illustrated in Section 3.

C

2.2 Damage evolution and constitutive model

AC

Once the above failure criterions are satisfied, the material stiffness of the damaged element will

degrade gradually based on the linear softening law [24,25], which the strength of the ply decreases

linearly with strain. The energy-based linear softening law is defined by effective stress and displacement

as shown in Fig. 1, wherein the fracture energy dissipating during the damage process is assumed to be

equal to the area under the effective stress- displacement curve:

5
 ACCEPTED MANUSCRIPT

∞
∫ σ eqd (δ eq ) = Gc (6)
0

In this equation σ eq and δeq are the effective stress and displacement respectively, which are given by

the following relations:

PT
For fiber failure:

δ eq = lcε11 , σ eq = lc (σ 11ε 11 ) / δ eq (7)

RI
For matrix failure:

SC
δ eq = lc ε 222 + ε122 + ε 232 , σ eq = lc (σ 22ε 22 + τ 12ε12 + τ 23ε 23 ) / δ eq (8)

where lc is the characteristic length of the element in ABAQUS [23]. For fiber failure, the fracture energy

U
is set to be equal to the fracture toughness (Gfc) of composites in the fiber direction. For matrix failure, the
AN
fracture energy is assumed to follow the mixed-mode fracture energy criterion, i.e. B-K law [26]:

Gs η
Gc = Gnc + (Gsc − Gnc )( ) (9)
M

Gn + Gs

1 1
Gn = (σ 22ε 22 )lc , Gs = (τ 12ε12 + τ 23ε 23 )lc (10)
D

2 2

where Gn and Gs are the normal and shear fracture toughness respectively, and η is a constant.
TE

Based on the linear softening law, the damage variable for each failure mode is defined by the effective
EP

displacement and can be calculated as the followings:

δ eq0 (δ eqf − δ eq )
d = 1− , δ eq0 ≤ δ eq ≤ δ eqf (11)
δ eq (δ eqf − δ eq0 )
C
AC

where δeq0 is the equivalent displacement when the damage initiates, and δeqf is the equivalent

displacement at which the material is completely failed.

Once the damage variables are obtained, the damaged constitutive relationship for 3-D composite

laminates can be expressed as [27,28]:

6
 ACCEPTED MANUSCRIPT

(1−df )C11 (1−df )(1−dm)C12 (1−df )(1−dm)C13 0 0 0 

 (1−dm)C22 (1−dm)C23 0 0 0 
 
 (1−dm)C33 0 0 0 
Cd =  
(12)
 (1−ds )C44 0 0 
 (1−ds )C55 0 
 
 sym (1−ds )C66 

PT
where df, dm and ds are damage variables calculated from Eq. (11), representing the fiber breakage, matrix

RI
cracking and shear failure respectively. Note that the shear failure is not independent and can be

considered as the interaction of matrix cracking and fiber breakage failure. Thus the damage variable

SC
relating to shear is given by:

d s = 1 − (1 − d f )(1 − d m ) (13)

U
AN
3. Micromechanics analysis

Though the real fiber distribution of composites are in random orders, the mechanical behaviors can be
M

successfully represented by those of hexagonal distribution, as found by Jin et al. [29] in a recent article.

Here, a hexagonal periodical representative volume element (RVE) with explicit fiber and matrix
D

constituent is adopted to model the real ply structure as shown in Fig. 2. Periodical and symmetric
TE

boundary conditions are specified for the unite cell to ensure the deformation compatibility or symmetric

conditions within the unit cell model. Materials properties used for the micromechanics analysis are listed
EP

in Table 1, where fibers are considered to be orthotropic and matrix is homogeneous and isotropic, the

fiber volume fraction is 0.62. Note that these constituent properties can be backcalculated from ply
C

properties or obtained directly from experiments.

AC

In order to obtain the mechanical and thermo-mechanical stress amplification factors, the unit cell is

given unit thermal load and unit mechanical load for different stress cases. Then the stress distribution at

each point of the unit cell can be regarded as the value of a certain entity of stress amplification factor for

that point [30]. The micro stress distributions of the unit cell model under various loading conditions are

shown in Fig. 3. For simplicity, some typical points in the unit cell are selected to extract the stress

7
 ACCEPTED MANUSCRIPT

amplification factors as shown in Fig. 2. The points F1 to F13 are located in the fiber and points M1 to

M17 are located in the matrix, they are used to calculate the values of amplification factors for fiber and

matrix respectively. It should be noted that, the suite of micromechanics analysis for a certain composite

structure needs to be performed only once, then the resulting amplification factors are stored and reused

PT
in the UMAT for stress amplification of macro-FE analysis.

RI
4. Finite element modeling and micromechanics-based progress failure analysis of composite vessel

4.1 Finite element model

SC
Fig. 4 shows the carbon fiber/epoxy composite hydrogen storage vessel with the inner diameter of 185

mm. Specific geometry parameters of the composite vessel are listed in Table 2. The vessel consists of a

U
6061-T6 aluminum liner and several outer T700 carbon fiber reinforced epoxy composite layers. The
AN
composite layers are assumed to be orthotropic and linear-elastic, while the aluminum liner is treated as

elasto-plastic material with the yield strength of 245 MPa. Mechanical properties of the materials are
M

listed in Table 1 and 3. It should be noted that, the uniaxial tensile and compressive tests of composite

laminates were carried out according with GB/T 1040.4-06, to get the longitudinal and transvers ply
D

strength. Once the ply strength are obtained, the constituent strength parameters can be back calculated
TE

from the ply strength using the described micromechanics unit cell model [31].

The FE mesh of the vessel is shown in Fig. 5, the model is supported on the front and loaded with
EP

internal pressure at the inner surface. The implicit analysis method in ABAQUS is adopted to solve the

nonlinear problem, which the internal pressure load is increased iteratively by the Newton-Raphson
C

algorithm at each load step. Besides, the nonlinear effects of large deformation are also considered. Total
AC

8,6120 3D linear brick elements are used in the analysis, the calculation results change little when further

refine the mesh. Note that there will be a significant temperature rise during the hydrogen refueling

process within the composite vessel. In order to consider the thermal effects on the mechanical property

of the composite vessel structure, the sequentially coupled thermal-stress analysis method is adopted,

which is performed by first solving a steady-state heat transfer problem, then reading the temperature

8
 ACCEPTED MANUSCRIPT

solution into the stress analysis at each load step as a predefined field. The boundary condition of heat

transfer moedel is specified based on our former work [32]: the heat convection coefficient at inner

surface of the vessel is set to 300 W/m2·K, gas temperature within vessel is 358 K and the outer surface

temperature is set as constant of 293 K. The schematic temperature distribution of the vessel structure is

PT
shown in Fig. 6.

RI
4.2 Solution algorithm

SC
The flow chart for impletmenting the micromechanics-based progressive failure analysis is described

in Fig. 7. Mechanical loads combined with thermal loads are applied on the composite vessel at each load

U
step, then the micro-stress of each constituent is obtained using stress amplification factors coded in the

UMAT. Based on the micro stress, the micromechanics-based failure criteria is applied independently to
AN
the fiber and matrix phases to judge whether the constituents in element fail or not. Once the failure has

been detected, stiffness degradation of the damaged element is performed based on the calculated damage
M

variable of each constituent. The calculations are performed continuously under the same load level until
D

there is no element failure anymore, then the load increases and the procedure repeats with the degraded
TE

properties to find the next critical constituent in the elements. Finally the ultimate strength of the

composite vessel is obtained when a sudden catastrophic failure occurs in the composite structure. In
EP

addition, the viscous stabilization method is adopted to help enhance the numerical convergence so as to

obtain the ultimate load capacity of the composite vessel effectively. The numerical calculations are
C

implemented on a high-performance computer with eight processors, and the calculation lasts for about
AC

10 hours.

4.3 Results and discussions

Fig. 8 shows stress distributions of the vessel under different internal pressures. The maximum stress

mainly appears at the cylinder part near the head. The stress are in non-linear relationships with internal

pressure, which indicates the non-linear mechanical behaviors of the composite vessel structure under

high pressure. This is mainly due to the plasticity of the aluminum liner and the stiffness degradation of

9
 ACCEPTED MANUSCRIPT

the failed carbon fiber/epoxy composites. The damage progression of the composite vessel is also

obtained from the proposed simulation strategy. Fig. 9 shows the evolution of fiber damage variable and

Figs. 10 shows the failure patterns of composite layer with increasing pressure. Matrix starts to fail at the

joint of the cylinder and head under the pressure of about 60 MPa, while fiber failure initiates at about 80

PT
MPa and mainly appears in the cylinder part of the vessel. The failed elements leads to stress

redistributions and severe stress concentrations on neighboring elements, which further results in

RI
continuous element failure in the composite laminates.

SC
The load-displacement curve of the vessel predicted by the model is shown in Fig. 11. The linear

deflecting phase of the curve indicates the appearance of matrix cracking damage, while the nearly flat

U
phase suggests a large number of fiber failed and the advent of strain softening stage. Fiber show brittle

fracture behavior and determine to a large extent the load capacity of the vessel structure, while matrix
AN
damage has a much lesser effect on the strength of the composite laminates. The failure strength of the

composite vessel predicted by the current model is 95.7 MPa, which is in good agreement with the result
M

of 95-98 MPa obtained by performing burst experiments as shown in Fig. 12. The burst experiment is
D

performed on a multifunctional pressure test system at room temperature (20°C). A high-pressure pump
TE

with the maximum working pressure 200 MPa is employed to fill the vessel. The medium of the burst test

system is clear water and the rate of pressurization is 0.2 MPa/s, the pressure increases almost linearly
EP

with filling time until suddenly drops to zero, indicating the burst failure of the vessel. The vessel

fractured at the cylinder part near the head as shown in Fig. 12 (b), which corresponds well with the
C

simulation results. The good agreement between the experiment and FEA validates the progressive failure
AC

analysis algorithm proposed in the current study.

In order to evaluate the thermal effects on failure strength of the composite vessel, the predicted result

has also been compared with that of without considering thermal load. As shown in Fig. 13, the ultimate

strength predicted by MMF is about 5% lower than that of without considering thermal effects, i.e. MMF-

Without . However the difference between the two MMF based simulations is relatively small, which

indicates that the thermal effect is very small and the failure of composite vessel is mainly caused by the

10
 ACCEPTED MANUSCRIPT

high-pressure load. Furthermore, results predicted by the proposed MMF method are also compared with

another two modeling strategies, which based on the integration of MPDM with traditional Tsai-Wu [33]

and Hashin [34] failure criteria. As shown in Fig. 13, the load-displacement curves predicted by the three

different methods show small difference, all around 90Mpa, with the result of MMF more closely

PT
resemble the experimental value. However, compared to traditional methods, the MMF shows clearly the

damage initiation in each constituent, and predictes a more reasonable damage evolution process in the

RI
fiber and matrix baesd on the consitituent propertity degradation. The proposed MMF method is more

SC
efficient and accurate to simulate the progressive failure process and predict the ultimate strength of the

composite vessel sturcture.

U
5. Conclusion
AN
A numerical algorithm for predicting complex failure mechanisms and progressive post-failure

behaviors of composite laminates is formulated based on micromechanics of failure (MMF) theory,

M

wherein the MMF is used to predict the failure initiation at constituent level and the MPDM is employed

to account for the post-initial failure behavior of the damaged materials. Furthermore, the
D

micromechanics analysis of a typical unit cell model is also performed, and the stress amplification
TE

factors which bridge the micro- and macro-level analysis are obtained. The model is applied to predict the

ultimate strength and complex failure behaviors of the composite vessel subjected to both thermal and
EP

mechanical loadings. The predictions of the model are also compared with those by experiment and

traditional finite element analysis, and reasonably good agreements are observed. This work provides
C

theoretical guidance for the safety and economical design as well as practical application of the composite
AC

vessel in fields of hydrogen fuel cell vehicles.

Acknowledgments

The research is supported by the Natural Science Foundation of Zhejiang Province, China (Number:

2012C24020).

11
 ACCEPTED MANUSCRIPT

References

[1] Vasiliev V V, Krikanov A A, Razin A F. New generation of filament-wound composite pressure vessels

for commercial applications[J]. Composite structures, 2003, 62(3): 449-459.

[2] Liu P F, Xing L J, Zheng J Y. Failure analysis of carbon fiber/epoxy composite cylindrical laminates

PT
using explicit finite element method[J]. Composites Part B: Engineering, 2014, 56: 54-61.

[3] Cox B, Yang Q. In quest of virtual tests for structural composites[J]. science, 2006, 314(5802): 1102-

RI
1107.

SC
[4] Yang J C. A thermodynamic analysis of refueling of a hydrogen tank[J]. international journal of

hydrogen energy, 2009, 34(16): 6712-6721.

U
[5] Galassi M C, Baraldi D, Iborra B A, et al. CFD analysis of fast filling scenarios for 70 MPa hydrogen
AN
type IV tanks[J]. International Journal of Hydrogen Energy, 2012, 37(8): 6886-6892.

[6] Zheng C, Wang L, Li R, et al. Fatigue test of carbon epoxy composite high pressure hydrogen storage
M

vessel under hydrogen environment[J]. Journal of Zhejiang University SCIENCE A, 2013, 14(6): 393-

400.
D

[7] Lin S, Jia X, Sun H, et al. Thermo-mechanical properties of filament wound CFRP vessel under
TE

hydraulic and atmospheric fatigue cycling[J]. Composites Part B: Engineering, 2013, 46: 227-233.

[8] Wang L, Zheng C, Luo H, et al. Continuum damage modeling and progressive failure analysis of
EP

carbon fiber/epoxy composite pressure vessel[J]. Composite Structures, 2015, 134: 475-482.
C

[9] Orifici A C, Herszberg I, Thomson R S. Review of methodologies for composite material modelling
AC

incorporating failure[J]. Composite Structures, 2008, 86(1): 194-210.

[10] Hinton M J, Soden P D. Predicting failure in composite laminates: the background to the exercise[J].

Composites Science and Technology, 1998, 58(7): 1001-1010.

[11] Soden P D, Hinton M J, Kaddour A S. A comparison of the predictive capabilities of current failure

theories for composite laminates[J]. Composites Science and Technology, 1998, 58(7): 1225-1254.

12
 ACCEPTED MANUSCRIPT

[12] Garnich M R, Akula V M K. Review of degradation models for progressive failure analysis of fiber

reinforced polymer composites[J]. Applied Mechanics Reviews, 2009, 62(1): 010801.

[13] Lee S Y, Roh J H. Two-dimensional strain-based interactive failure theory for multidirectional

composite laminates[J]. Composites Part B: Engineering, 2015, 69: 69-75.

PT
[14] Gosse J H, Christensen S. Strain invariant failure criteria for polymers in composite materials[J].

RI
AIAA, 2001, 1184: 11.

[15] Garnich M R, Hansen A C. A multicontinuum theory for thermal-elastic finite element analysis of

SC
composite materials[J]. Journal of Composite Materials, 1997, 31(1): 71-86.

[16] Ha S K, Jin K K, Huang Y. Micro-mechanics of failure (MMF) for continuous fiber reinforced

U
composites[J]. Journal of Composite Materials, 2008.
AN
[17]Crouch R D, Clay S B, Oskay C. Experimental and computational investigation of progressive

damage accumulation in CFRP composites[J]. Composites Part B: Engineering, 2013, 48: 59-67.
M

[18] Ridha M, Wang C H, Chen B Y, et al. Modelling complex progressive failure in notched composite
D

laminates with varying sizes and stacking sequences[J]. Composites Part A: Applied Science and
TE

Manufacturing, 2014, 58: 16-23.

[19] Barbero E J, Lonetti P. An inelastic damage model for fiber reinforced laminates[J]. Journal of
EP

Composite Materials, 2002, 36(8): 941-962.

[20] Lapczyk I, Hurtado J A. Progressive damage modeling in fiber-reinforced materials[J]. Composites

C

Part A: Applied Science and Manufacturing, 2007, 38(11): 2333-2341.

AC

[21] Wang C, Liu Z, Xia B, et al. Development of a new constitutive model considering the shearing

effect for anisotropic progressive damage in fiber-reinforced composites[J]. Composites Part B:

Engineering, 2015, 75: 288-297.

[22] Tay T E, Liu G, Yudhanto A, et al. A micro-macro approach to modeling progressive damage in

composite structures[J]. International Journal of Damage Mechanics, 2007.

13
 ACCEPTED MANUSCRIPT

[23] Abaqus Analysis User’s Manual, Version 6.10; 2011.

[24] Camanho P P, Dávila C G. Mixed-mode decohesion finite elements for the simulation of

delamination in composite materials[J]. 2002.

[25] Li G, Li C. Assessment of debond simulation and cohesive zone length in a bonded composite

PT
joint[J]. Composites Part B: Engineering, 2015, 69: 359-368.

RI
[26] Benzeggagh M L, Kenane M. Measurement of mixed-mode delamination fracture toughness of

unidirectional glass/epoxy composites with mixed-mode bending apparatus[J]. Composites Science

SC
and Technology, 1996, 56(4): 439-449.

[27] Highsmith A L, Reifsnider K L. Stiffness-reduction mechanisms in composite laminates[J]. Damage

in composite materials, 1982, 775: 103-117.

U
AN
[28] Tserpes K I, Papanikos P, Kermanidis T. A three-dimensional progressive damage model for bolted

joints in composite laminates subjected to tensile loading[J]. Fatigue & Fracture of Engineering
M

Materials & Structures, 2001, 24(10): 663-675.

D

[29] Jin K K, Oh J H, Ha S K. Effect of fiber arrangement on residual thermal stress distributions in a

TE

unidirectional composite[J]. Journal of composite materials, 2007, 41(5): 591-611.

[30] Sun X S, Tan V B C, Tay T E. Micromechanics-based progressive failure analysis of fibre-reinforced

EP

composites with non-iterative element-failure method[J]. Computers & Structures, 2011, 89(11):

1103-1116.
C

[31] Cai H, Miyano Y, Nakada M, et al. Long-term fatigue strength prediction of CFRP structure based on
AC

micromechanics of failure[J]. Journal of composite materials, 2008, 42(8): 825-844.

[32] Wang L, Zheng CX, Li R, Chen BB, Wei ZX. Numerical analysis of temperature rise within 70MPa

composite hydrogen vehicle cylinder during fast refueling[J]. Journal of Central South University

2014; 21(7): 2772-8.

[33] Tsai S W, Wu E M. A general theory of strength for anisotropic materials[J]. Journal of composite

14
 ACCEPTED MANUSCRIPT

materials, 1971, 5(1): 58-80.

[34] Hashin Z. Failure criteria for unidirectional fiber composites[J]. Journal of applied mechanics, 1980,

47(2): 329-334.

PT
RI
Figure Captions

U SC
AN
M

Fig. 1. Energy based linear softening law.

D
TE
C EP
AC

Fig. 2. Hexagonal unit cell model and selected points for extraction of amplification factors.

15
 ACCEPTED MANUSCRIPT

PT
RI
U SC
Fig. 3. Micro stress distributions in the unit cell due to unit mechanical and thermal loading.
AN
M
D
TE

Fig. 4. Carbon fiber/epoxy composite hydrogen storage vessel.

C EP
AC

Fig. 5. Mesh model of the vessel.

16
 ACCEPTED MANUSCRIPT

PT
RI
SC
Fig. 6. Schematic temperature distribution of the vessel (unit: K).

U
AN
M
D
TE
C EP
AC

Fig. 7. Flow chart for implementing the micromechanics-based progress failure analysis of composite

vessel.

17
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Fig. 8. Distributions of the Mises stress (unit: MPa) at the internal pressure of (a) 60 MPa, (b) 70 MPa, (c)

80 MPa and (d) 90 MPa.

M
D
TE
C EP
AC

Fig. 9. Evolution of fiber damage variable at the internal pressure of (a) 75 MPa, (b) 80 MPa, (c) 85 MPa

and (d) 90 MPa.

18
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Fig. 10. Failure patterns of composite layer at the internal pressure of (a) 60 MPa, (b) 70 MPa, (c) 80 MPa

and (f) 90 MPa.

M
D
TE
C EP
AC

Fig. 11. Predicted load-displacement curve of the vessel compared with experimental burst pressure.

19
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Fig. 12. Burst experiment of the composite vessel: (a) vessel specimen under experiment, (b) vessel
M

specimen after burst.

D
TE
C EP
AC

Fig. 13. Comparison of results predicted by different methods.

20
 ACCEPTED MANUSCRIPT

Tables

Table 1. Material properties of the composite vessel.

Material properties Ply Fiber Matrix Aluminum

PT
(T700/Epoxy) (T700) (Epoxy) (6061-T6)
E1 (GPa) 142 232 3.5 70
E2= E3 (GPa) 8.5 18 3.5 70
G12= G13 (GPa) 3.7 8.7 1.25 26.3

RI
G23 (GPa) 2.6 5.8 1.25 26.3
v12= v13 0.25 0.2 0.35 0.33
v23 0.42 0.49 0.35 0.33

SC
a1 (10-6/K) 0.1 0 58 23.5
a2= a3 (10-6/K) 22.5 12 58 23.5

Table 2. Geometry parameters of the composite vessel.

Material Layer

U Layer angle
(o)
Layer thickness
AN
(mm)
Aluminum (6061-T6) Liner — 2.5
1 90 2.1
2 12 0.87
M

3 15 0.87
Composites 4 19 0.87
(T700/Epoxy) 5 90 2.1
6 22 0.87
D

7 27 0.87
8 32 0.87
TE

9 38 0.6
10 90 0.54
EP

Table 3. Constituent strength parameters.

Properties Constituent (T700/Epoxy)
Longitudinal tensile strength Tf (MPa) 4150
Longitudinal compressive strength Cf (MPa) 2075
C

Transverse tensile strength Tm (MPa) 105

Transverse compressive strength Cm (MPa) 241
AC

Longitudinal toughness Gfc (N/mm) 106

Transverse normal toughness Gnc (N/mm) 0.28
Transverse shear toughness Gsc (N/mm) 0.79

21