Professional Documents
Culture Documents
Fatigue Analysis of A Composite Ring: Experimental and Theoretical Investigations
Fatigue Analysis of A Composite Ring: Experimental and Theoretical Investigations
COMPOSITE
Article M AT E R I A L S
Journal of Composite Materials
0(0) 1–14
! The Author(s) 2020
Fatigue analysis of a composite Article reuse guidelines:
sagepub.com/journals-permissions
ring: Experimental and DOI: 10.1177/0021998320925163
journals.sagepub.com/home/jcm
theoretical investigations
Abstract
The behavior of composite rings against fatigue loading is analyzed. Experimental study is performed to evaluate the
ultimate hoop strength of the composite ring as a requirement before fatigue testing. Then, fatigue tests are performed at
three different load levels. Afterward, progressive damage modeling in the context of continuum damage mechanics is
utilized to theoretically estimate fatigue lifetimes of the investigated rings. The modeling consists of three parts as stress
analysis, damage estimation and degradation of mechanical properties. It is mainly aimed to reduce the required runtime
for the accomplishment of fatigue analysis while the accuracy is at least compromised. Therefore, a manual technique is
adapted for extracting stress components on the ring. Degradation of mechanical properties takes place either through
gradual procedure or sets of sudden laws. The gradual degradation of mechanical properties is performed based on the
damage progression as a function of cycle numbers and stress state. Sudden degradation is done after occurrence of
failure. After reduction of mechanical properties, the model is updated and whole procedure repeats till the whole layers
fail. Finally, the results of fatigue modeling are compared with performed experimental study and a very good agreement
is observed.
Keywords
Composite ring, fatigue analysis, experimental study, progressive damage modeling, damage evolution
based on the stiffness reduction arising from adhesive Table 1. Mechanical properties of utilized glass fiber and epoxy
damage.18 Rafiee and Eslami employed progressive resin.
damage modeling (PDM) for estimating the fatigue
Mechanical properties Glass fiber Epoxy resin
life of composite pipes under cycling internal pres-
sure.19 The damage evolution was captured using stiff- Young’s modulus (GPa) 81 3.5
ness degradation technique and Taguchi method was Shear modulus (GPa) 32 1.25
utilized to obtain proper degradation pattern.19 The Poisson’s ratio 0.21 0.35
developed modeling procedure in Rafiee and Eslami19 Strength (MPa) 2000 Tensile: 75,
was extended to estimate fatigue lifetime of composite compressive: 250
pipe under cyclic internal pressure with variable ampli- Density (g/cm3) 1.15 2.56
tudes.20 Unlike the normal practice, stochastic fatigue
modeling was conducted treating maximum stress amp-
litude, mean stress amplitude and number of cycles as
random variables resembling random fatigue loading
pattern.20 Shabani et al.21 employed a PDM technique
for extracting fatigue lifetime of a notched composite
ring. A complete set of stiffness and strength degrad-
ation rules were obtained experimentally and fed into
the model.
This research deals with experimental determination
and theoretical estimation of fatigue life in notched
composite rings cut from glass/epoxy filament wound
tubes. The remaining of the paper is organized as fol-
Figure 1. Filament winding process and produced tubes.
lows. Firstly, experimental fatigue study is carried out
on the composite rings at different load levels. Then,
fatigue simulation is conducted to predict the fatigue life. Twelve ring specimens were cut from produced tube
The employed fatigue modeling and its involved stages with the width of 25.4 mm. Circular notches were pro-
are explained in detail. The employed fatigue modeling vided in two areas placed 180 apart with the radius of
procedure is categorized under PDM group where stiff- 5.2 mm in accordance with reflected dimensions in
ness degradation laws account for damage evolution in ASTM D2290.22 As a result, the distance between the
the context of continuum damage mechanics. Aiming to two notches in the test ring is 15 mm. The geometrical
reduce the required runtime for the accomplishment of dimensions of the ring specimens are shown in Figure 2.
fatigue analysis as the main streamline of this research,
the main components of the modeling is simplified while
the accuracy is at least compromised.
Measuring hoop tensile strength
Prior to performing the fatigue test, it is necessary to
determine the hoop tensile strength (HTS) of the rings.
Experimental study The procedure A reflected in ASTM D2290 is used to
determine the tensile HTS of the rings.22 Known as
Materials and test specimen split-disk method, the test is performed using an uni-
A glass/epoxy tube with inner radius of 50.8 mm and versal testing machine equipped with split-disk test fix-
total length of 6000 mm was produced using reciprocal ture as shown in Figure 3.
filament winding process. The mechanical properties of The ring samples were mounted on the split-disk
the utilized glass fiber and epoxy resin are listed in while the reduced areas are placed at the split in the
Table 1. The layup configuration of produced tube fixture. Adjusting the rate of the cross-head movement
was [543] with the total thickness of 2.8 mm. at 2 mm/min, the tensile load was applied to the ring
Aforementioned winding angles are measured from and increased until the failure was observed. The test
the axial direction of the tube. The produced tube was performed on three different ring specimens and
was cured 80 C for 4 h and then post-cured at 120 C recorded data are presented in Table 2.
for another 4 h. The filament winding procedure and
also produced tube is shown in Figure 1.
Conducting fatigue test
The utilized glass fiber is E386T-E6 direct roving
produced by Jushi company and Araldite LY556 The same testing apparatus used for measuring HTS in
epoxy resin supplied by Huntsman company is used the previous section was used for the purpose of fatigue
as the resin system. testing. A cyclic tensile-tensile load with constant
Rafiee et al. 3
Theoretical modeling
Theoretical modeling of fatigue is conducted in this
research using PDM technique. Integrating failure cri-
teria, fatigue life models and material degradation
Figure 2. Geometrical dimensions of the ring specimen. rules, PDM techniques simulate the fatigue damage ini-
tiation and propagation.
Under cyclic loading conditions, the gradual accu-
mulation of fatigue damage can result in a final failure
of the composite parts. In fact, fatigue damage occurs
cycle by cycle, and as the number of cycles increases,
the damage accumulates and growths until that com-
ponent is no longer able to tolerate any more damage
and thus failure occurs. Due to the complexity of fati-
gue damage mechanisms in composites, it is not pos-
sible to directly measure minor damages such as very
small strains and micro-cracks. As an alternative
method, the damage evolution is indirectly monitored.
Namely, the influence of minor damages on physical
quantities such as stiffness and strength can be mea-
sured and used to define damage variables. This strat-
egy of fatigue modeling falls into the category of PDM
techniques.
The procedure of the employed PDM technique in
this research consists of three main parts as stress ana-
lysis, damage estimation and degradation of properties.
Figure 3. Split-disk test assembly. The stresses induced in all layers are firstly analyzed.
Afterwards, the stresses are applied to a series of
amplitude and stress ratio of 0.1 was applied to the ring damage relationships to calculate the damage variable
specimens and the number of cycles to failure was in each constitutive plies, individually. The material
counted and recorded in each test. The experiment properties of each ply are gradually degraded based
was performed at a frequency of 10 Hz avoiding over- on the evolution of the damage variable with respect
heating. The maximum of applied stress is chosen as to the number of cycles. Furthermore, when failure
60%, 70% and 80% of the HTS measured in the pre- occurs in any layer, its properties are suddenly
ceding section. For each loading scenario, three ring degraded. For this purpose, the model should be able
specimens were tested. The recorded data of fatigue to simulate the cycle-to-cycle reduction of the proper-
test are presented in Table 3. ties of the Uni-Directional (U-D) ply under the multi-
As it can be seen from presented results in Table 3, axial fatigue loading. As the number of cycles increases,
decreasing the level of maximum applied stress, the stress analysis is performed for the material with
4 Journal of Composite Materials 0(0)
Aa A B B Minimum area
Width Thickness Width Thickness of reduced Force Strengthc
(mm) (mm) (mm) (mm) sectionb (mm2) (KN) (MPa)
80% 15.026 2.786 14.876 2.71 82.233 24.553 198 239 290 242.33 46.09
HTS
70% 15.193 2.776 14.836 2.813 83.933 21.93 432 513 536 493.66 54.62
HTS
60% 15.21 2.77 15.11 2.88 85.663 19.176 680 777 1349 935.3 361.5
HTS
HTS: hoop tensile strength.
Figure 4. Fatigue testing machine and failed specimen after fatigue test.
updated properties and the whole process is repeated the geometry and loading conditions of the
until the ultimate failure of the part occurs and the investigated ring, the stress components can be
specimen can no longer bear any more load. The extracted accurate enough using finite element model-
flowchart of the PDM technique used in this study ing. Therefore, at each stage of fatigue modeling,
for simulating fatigue on the composite ring is shown the finite element model should be called. Fatigue
in Figure 5 and involved steps are described in the fol- modeling will therefore be very time consuming
lowing sections in details. and tedious. In this section, in order to reduce the mod-
eling time, the manual calculation method based on
classical lamination theory (CLT) is employed to
Stress analysis
derive the stress components as an alternative to the
Stress components on each ply are required to be finite element method. In other words, it is mainly
calculated at each stage of cyclic loading. Due to aimed to reduce the required runtime for the
Rafiee et al. 5
Figure 5. Flowchart of fatigue modeling through PDM technique. PDM: progressive damage modelling.
Ef
1 þ 2Vf 1
Ey ¼ Em , ¼ EEmf
1 Vf þ2
Em
Gf
1 þ s Vf 1
Es ¼ Gm , s ¼ GGmf ð4Þ
1 s Vf þ1
Gm
It is also worth noting the moment is induced by the max ¼ Kt nom ¼ 1:66 nom ð10Þ
change in contour of the ring between the two disk
Rafiee et al. 7
Figure 7. Graphical demonstration of obtaining on-axis stress components from off-axis strains.
Comparing the far-field stress with the value specimen was flat, the load magnification factor would
obtained using described manual method, they not be required to be defined and nom would be con-
obtained a load magnification factor as below27 sidered the same as CLT.
As it can be seen, if the stress obtained by the
nom ¼ KL CLT ¼ 1:63 CLT ð11Þ manual calculation method is multiplied by 2.7, the
maximum stress will be obtained in the neighborhood
Due to this fact that the effect of specimen curvature of the notch. So, the described method used in this
is not taken into account in CLT, the load magnifica- study will overestimate the stress values presenting a
tion factor is adopted. In other words, if the analyzed conservative approach.
8 Journal of Composite Materials 0(0)
. U-D layer under transverse fatigue loading . U-D layer under shear fatigue loading condition
conditions
4 3 2
D~ ¼ 335:25N~ 1054:9N~ þ 1239:4N~ 643:45N~
2m1
fð, ult Þ ¼ ð1 SÞ ð1 þ 3:1mCÞ ð15Þ
þ 124:73 ð22Þ
. U-D layer under shear fatigue loading conditions
fð, ult Þ ¼ 2:79Sm ð16Þ The number of normalized cycles (N) ~ reflected in
equations (17) to (22) is calculated by the following
equation28
In these equations, S represents the ratio of max-
imum applied stress to the ultimate strength of material logðnÞ logð0:25Þ
N~ ¼ ð23Þ
( max = ult ) and m is S(S-1). log Nf logð0:25Þ
The amount of the normalized damage D~ is related
to the number of normalized cycles N. ~ Examining the where n denotes the number of applied cycles and Nf is
available experimental data shows that from the begin- the number of cycles to failure. Nf is obtained for a U-
ning of cyclic loading (N~ ¼ 0) to N~ ¼ 0:67, the normal- D layer under the desired uniaxial stress state and stress
ized damage grows slowly and almost linearly and then ratio employing the following equation28
increases with higher rate and non-linearly until the
failure occurs (N~ ¼ 1).28 So, this trend can be divided lnða=fÞ
¼ A þ B log Nf ð24Þ
into two phases. Normalized damage equations in lnð1 qÞðC þ qÞ
terms of the number of normalized cycles for different
loading states in the first phase (0 N ~ 0:67) are q ¼ m = t
28
expressed as below: ð max þ min Þ
m ¼
2 ð25Þ
. U-D layer under longitudinal fatigue loading
a ¼ a = t
condition
ð max min Þ
a ¼
D~ ¼ 0:1493N~ ð17Þ 2
c =t Longitudinal loading
. U-D layer under transverse fatigue loading condition C¼ ð26Þ
t =c Transverse loading
D~ ¼ 0:0361N~ ð18Þ
max and min are the maximum and minimum stres-
. U-D layer under shear fatigue loading condition ses created in the desired U-D layer. t and c stand
for tensile and compressive strengths, respectively.
D~ ¼ 0:3983N~ ð19Þ Equation (24) is valid for tension and compression
modes of loading and for the case of shear loading, it
is rearranged as below28
For the second phase (0:67 N~ 1), the relations
are no longer linear and are in the form of 4th-order lnða=fÞ
log ¼ A þ B log Nf ð27Þ
equations as below:28 ln½ð1 qÞðC þ 1Þ
. U-D layer under longitudinal fatigue loading condition Parameters A, B and f are curve fitting parameters
obtained from the experimental data. These values for
4 3 2 the same glass/epoxy composites used to fabricate the
D~ ¼ 5:788N~ 6:8785N~ þ 2:4153N~ 0:2654N~
ð20Þ investigated rings in this research are obtained by
þ 0:0061
Rafiee and Eslami19 using Taguchi technique and
reported in Table 4. Since reported normalized life
. U-D layer under transverse fatigue loading condition curves in previous equations belong to carbon/epoxy
composites, they have modified aforementioned
4 3 2 curves for the same glass/epoxy composites used in
D~ ¼ 242:72N~ þ 715:04N~ 768:3N~ þ 358:19N~
this study in a manner to obtain minimum difference
61:22 ð21Þ
between theoretical prediction and experimental data.19
In other words, a generic behavior is estimated for
10 Journal of Composite Materials 0(0)
Loading type B A
stress level in the layers also decreases. Therefore, it is A comparison of the estimated fatigue lifetimes and
very probable that a situation happens while a very measured ones through experimental study is presented
weak material is exposed to a very low level of stress. in Figure 11.
Under this circumstance, the introduced method under It is observed that the present model is able to simu-
section Sudden degradation of mechanical properties for late the fatigue behavior of the specimen properly and
identifying the occurrence of failure is not able to dis- estimate the fatigue lifetime accurately enough.
tinguish failure and thus the code will never stop. The Moreover, the lifetimes in all cases are underestimated
ultimate static strain is used as another termination cri- attributed to the conservative nature of the modeling
terion to overcome this difficulty. At each step, the because of the overestimation of stress analysis.
obtained strain from the analysis is compared with Considering experimentally measured 1349 cycles at
the static failure strain ("ult ) and if it exceeds that the load level which is 60% of HTS loading as scattered
value, the program is terminated and corresponding data, the modeling is slightly overestimated the fatigue
number of cycles is reported as the fatigue life. The lifetime for this case as an exception.
value of the stress at the moment of failure is the The reduction trends of longitudinal, transverse and
same as the ultimate strength ðult Þ. The ultimate shear moduli are shown in Figure 12 for a ply of the
strength is calculated using Tsai-Wu failure criterion.24 investigated ring in all loading statuses. Comparing all
The required initial strength components for Tsai-Wu the trends, it is evident that the shear modulus
criterion are calculated in accordance with Appendix 2. decreases significantly and becomes zero, while both
Tsai-Wu criterion is an interactive rule taking into longitudinal and transverse moduli slightly decrease.
account the interaction of different stress components The sudden degradation of shear modulus to zero
and also categorized as a simple and fast rule. implies on the occurrence of shear failure mode, while
Moreover, Tsai-Wu criterion is conservative enough the other failure modes are not experienced.
in comparison with other single-criterion rules for com- Monitoring the trend of strength components also con-
posite materials. Lastly, the ultimate strength is firms the occurrence of shear failure mode, since just
required to be calculated and identification of the fail- shear strength is suddenly degraded to zero.
ure mode is not concerned in this section and Tsai-Wu It is worth noting that in the specific layup of the
failure criterion meets the purpose. investigated composite ring, the modeling is terminated
by the static failure strain criterion at all applied load
levels. For all stress levels, the values of stiffness are
Results and discussion
reduced to such a degree that the matrices reflected in
The described model in the previous section is imple- equation (1) become very small matrices. Thus,
mented on the composite rings examined in the experi- obtained strain values are very large in a manner that
mental study section and the results are compared. overall strain exceeds the static failure strain.
Figure 12. The degradation trends of longitudinal, transverse and shear moduli.
The runtime of the fatigue modeling is about 30 min not suffer from stiffness loss anymore and as a subse-
on a computer powered with Core i7-7500U processor quent fatigue failure is identified by the introduced
running at 2.7 GHz and 12-GB memory. static failure criterion. Therefore, it is believed that
As it was extensively explained, the utilized model- that the model is limited to the LCF regime, since
ing procedure is carried out replying solely on the strength degradation is not taken into account.
basis of stiffness degradation. The load is transferred Therefore, the application of explained approach oper-
from the layers suffering from stiffness loss to the stif- ating solely on the stiffness degradation is limited to
fer layers and thus a slower degradation of stiffness is LCF cases and its performance for HCF regime is
pertinent to the weaker layers. The weaker layers do required to be examined.
Rafiee et al. 13
12. Uyaner M, Kara M and Şahin A. Fatigue behavior of 31. Chamis CC. Simplified composite micromechanics equa-
filament wound E-glass/epoxy composite tubes damaged tions for strength, fracture toughness and environmental
by low velocity impact. Composites Part B 2014; 61: effects, Technical Memorandum N84–27832, NASA,
358–364. Lewis Research Center, Cleveland, Ohio, USA, 1984.
13. Krishnan P, Abdul Majid MS, Afendi M, et al. Effects of
winding angle on the behaviour of glass/epoxy pipes
under multiaxial cyclic loading. Mater Design 2015; 88:
196–206. Appendix 1
14. Eggers F, Almeida JHS, Azevedo CB, et al. Mechanical Calculated mechanical properties
response of filament wound composite rings under ten-
sion and compression. Polym Test 2019; 78: 105951.
15. Knox EM, Cowling MJ and Hashim SA. Fatigue per- Ex ¼ 50½GPa, Ey ¼ 15:26½GPa,
ð33Þ
formance of adhesively bonded connections in GRE Es ¼ 4:36½GPa and x ¼ 0:26
pipes. Int J Fatigue 2000; 22: 513–519.
16. Kara M and Kırıcı M. Effects of the number of fatigue 2 3
52:03 37:48 0
cycles on the impact behavior of glass fiber/epoxy com- 6 7
posite tubes. Compos Part B Eng 2017; 12315: 55–63. A ¼ 4 37:48 82:70 0 5 106 ½N=m ð34Þ
17. Taşyürek M and Tarakçioğlu N. Enhanced fatigue behav- 0 0 38:09
ior under internal pressure of CNT reinforced filament
wound cracked pipes. Compos Part B Eng 2017; 1241:
2 3
0 0 3:53
23–30. 6 7
18. Kim JS, Bae KD, Lee C, et al. Fatigue life evaluation of B¼4 0 0 7:45 5 103 ½N ð35Þ
composite material sleeve using a residual stiffness model. 3:53 7:45 0
Int J Fatigue 2017; 101: 86–95.
2 3
19. Rafiee R and Eslami F. Theoretical modeling of fatigue 33:89 24:42 0
phenomenon in composite pipes. Compos Struct 2017; 6 7
D ¼ 4 24:42 53:88 0 5 ½Nm ð36Þ
1611: 256–263.
20. Rafiee R. Stochastic fatigue analysis of glass fiber rein- 0 0 24:51
forced polymer pipes. Compos Struct 2017; 167: 96–102.
21. Shabani P, Taheri-Behrooz F, Maleki S, et al. Life pre-
diction of a notched composite ring using progressive Appendix 2
fatigue damage models. Compos Part B Eng 2019;
16515: 754–763. Required static strengths
22. ASTM D 2290-04. Standard test methods for apparent The required static strengths of each ply are calculated
tensile strength of ring or tubular plastics and reinforced
using below empirical formulation31
plastics by split disk method. American Society for Testing
and Materials, Philadelphia, 2004.
23. ASTM D 3171-06. Standard test methods for constituent
Em
XT ¼ Xf Vf þ Vm ð37Þ
contents of composite materials. American Society for Ef
Testing and Materials, Philadelphia, 2006.
24. Tsai SW, Hoa SV and Gay D. Composite materials, YT ¼ Vm Xm ð38Þ
design and applications. Boca Raton: CRC Press, 2003.
25. Gibson RF. Principles of composite material mechanics, XC ¼ 0:5XT ð39Þ
2nd ed. Boca Raton: CRC Press, 2007.
26. Rafiee R. Experimental and theoretical investigations on YC ¼ Vm X0m ð40Þ
the failure of filament wound GRP pipes. Compos Part B
Eng 2013; 45: 257–267. where Xf, Xm and X0m stand for fiber strength, matrix
27. Rafiee R and Abbasi F. Numerical and experimental ana- tensile strength and matrix compressive strength,
lyses of the hoop tensile strength of filament wound com-
respectively, reported in Table 1.
posite tubes. Mech Compos Mater2020; Accepted.
Employing equations (37) to (40), following values
28. Shokrieh MM and Lessard LB. Progressive fatigue
damage modeling of composite materials, Part I:
are obtained for strength components
Modeling. J Compos Mater 2000; 43: 1056–1080.
29. Shokrieh MM and Zakeri M. Generalized technique for XT ¼ 1234:6 ½MPa, XC ¼ 617:3 ½MPa,
ð41Þ
cumulative damage modeling of composite laminates. YT ¼ 30 ½MPa, YC ¼ 100 ½MPa
J Compos Mater 2007; 41: 2643–2656.
30. Rafiee R. Stochastic fatigue analysis of glass fiber rein- The in-plane shear strength is assumed 80 MPa
forced polymer pipes. Compos Struct 2017; 167: 96–102. reported by the resin supplier.