Accepted Manuscript

Structural optimization for CFRP cryogenic tank based on energy release rate

Hayato Fukui, Akinori Yoshimura, Ryousuke Matsuzaki

PII: S0263-8223(16)30789-9
DOI: http://dx.doi.org/10.1016/j.compstruct.2016.06.004
Reference: COST 7526

To appear in: Composite Structures

Received Date: 24 February 2016

Accepted Date: 1 June 2016

Please cite this article as: Fukui, H., Yoshimura, A., Matsuzaki, R., Structural optimization for CFRP cryogenic
tank based on energy release rate, Composite Structures (2016), doi: http://dx.doi.org/10.1016/j.compstruct.
2016.06.004

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.
 Structural optimization for CFRP cryogenic tank based on energy release rate
Hayato Fukui1, Akinori Yoshimura2, Ryousuke Matsuzaki1
1
Depertment of Mechanical Engineering, Tokyo University of Science, 2641 Yamazaki, Noda, Chiba
278-8510, Japan
2
Aeronautical Technology Directorate, Japan Aerospace Exploration Agency (JAXA), 6-13-1 Osawa,
Mitaka, Tokyo 181-0015, Japan

ABSTRACT
In the present study, we conducted structural optimization for CFRP cryogenic propellant tank.
Under the cryogenic temperature, large thermal stress between metallic material and CFRP might
cause the debonding failure. Optimizations were performed by considering crack propagation. Energy
release rate of debonding crack calculated by virtual crack closure method was used as objective
function. Two analytical conditions were considered: when merely the temperature of tank was
decreased and when the pressure was applied to the tank in the decreased temperature. Two types of
optimizations were conducted for each condition. One is liner wall thickness optimization. Another
optimization is curved wall shape optimization. On the basis of these results, we proposed a new
design of the CFRP cryogenic tank. The new design that has notch and optimized dorm shape can
reduce energy release rate near the cap and prevent crack propagation.
1. INTRODUCTION
For future reusable space transportation system, it is necessary to reduce structural weight in order to
improve its performance. Especially, propellant tank occupies a large proportion of structural weight. Thus
reduction of the weight of tank is essential factor to reduce structural weight drastically. For that reason,
application of carbon fiber reinforced plastics (CFRP) for propellant tank can reduce the weight of tank and
structural weight.
For pressure vessels and tank structures that are used in normal temperature and high pressure, the
methods to decide shape and parameters of filament winding for the application of composite materials and
fracture behavior are deeply discussed. Then, pressure vessels and tank structures that are made by FRP are
increasing. In the research about shape determining methods, there are Isotensoid surface[1] by based on netting
theory, method that is based on fracture law[2] and that consider the friction between mold and fiber[3, 4]. In
addition, fiber orientation has large effects on the performance of pressure vessel. For that reason, optimizations
for fiber orientation against maximum burst pressure[5], number of windings of helical and hoop layer, fiber
orientation at cylinder part and thickness of liner[6, 7] were conducted. Optimization method to determine fiber
orientation by using formulation that consider classical laminate theory was also researched[8]. Recently,
because of improvement of computer, optimal tank structures by based on various optimization methods[9-11]
are proposed. For example, Xu et al. [12, 13] conducted weight minimization for composite pressure vessel
under constraint conditions of burst pressure by using Adaptive Genetic Algorithm (AGA). Vafaeesefat[14]
optimized type 3 tanks by using adaptive response surface method (ARAM) and genetic algorithm. Effects of
parameter of filament winding on quality and durability of pressure vessel[15] and first-ply failure of composite
pressure vessel [16] were discussed. Progressive failure analysis[17, 18] which used probabilistic analysis, such
as Monte Carlo simulation, and Weibull distribution and structure[19] which can reduce energy release rate at
adhesion near the cap were also researched. From the above, we can conclude that researches about durability
and weight of usual composite pressure vessel are the main concern [5, 8, 12, 15-18, 20].
In comparison to usual high pressure vessels which are being applied to FRP, propellant tanks for space
transportation systems are required lower pressure than usual tanks. While, liquid propellants are cryogenic
liquid hydrogen and oxygen. For that reason, propellant tanks are exposed to cryogenic temperature in operation.
In cryogenic temperature, large amount of thermal stress and delamination between metal material and CFRP
would occur due to mismatch of these coefficients of thermal expansion (CTE)[21]. Application of invar alloy,
which CTE is nearly similar to CFRP, can prevent delamination. However, invar alloy is heavier and more
expensive than aluminum alloy. Thus, new structure of tank, which liner is made of lightweight and low cost
materials and which can prevent delamination, is required. In researches of composite tank in cryogenic
temperature, pressurized test under liquid nitrogen [22], measurements of strain by using FBG sensor in
Reusable Vehicle Testing (RVT) and burst tests[23] for composite pressure vessels in cryogenic temperature
were conducted by using actual tank structures. In the view of progressive failure, Ju et al. [24] discussed
progressive failure of composite tanks on the basis of classical laminate theory. There are few discussions about
tank structures and methods to determine the shape. Optimization for structural weight and bending modulus of
whole structure [25] was conducted and the structure ,which can prevent crack propagation by using aluminum
honeycomb[26], was proposed. However, the optimal structure, which can prevent delamination between
different materials, is not still proposed.
In this study, we conducted optimization to prevent crack propagation for cryogenic CFRP tank on the
basis of energy release rate. We used thickness of liner and structural shape for design valuables and proposed
the new structure, which can prevent delamination. In analyses, in order to consider maximum operating
pressure and temperature, it is necessary to consider the change of material properties due to change of
temperature. Thus, behaver of crack propagation in cryogenic temperature based on DCB, ENF and MMB tests
was considered in analyses. It is the biggest advance from previous studies.
2. Methods to obtain energy release rate using VCCT
In this study, in order to calculate energy release rate, we employed the virtual crack closure technique
(standard VCCT) [27]. The energy release rate G under constant loading condition can be calculated as,
U
G= (1)
A
where U and A denote strain energy and area of crack propagation, respectively. Equation (1) means that energy
release rate is equal to change of elastic energy when small crack propagation occurred. In this study, we
consider a crack in elastic media. The change of elastic energy between the state of before crack propagation
(Fig.1 (a)) and the state of after crack propagation (Fig.1 (b)) is equal to the work of the external force f (=
(fx,fy,fz)) which closes the crack. Then the work for closing crack W is expressed as,
uia + a
W = f i dui dA (2)
A 0
where uia+a represents relative displacement of upper and lower surfaces of crack. Because it is equal to the
change of elastic energy U, energy release rate under constant loading is expressed as,
W
G= (3)
A
Then we apply it to finite element method. When we assume that fi is nodal force and displacement i is relative
displacement at adjacent node P, the energy release rate G is represented as,
W 1
G= = f i i (4)
A 2A
where the nodes to calculate relative displacement i is deferent from the node to calculate nodal force (see
figure 2). In the standard VCCT method, it is assumed that when the node Q, which calculate nodal force, is
divided into two nodes, the crack opening displacement at Q will be equal to that of the node P before separation
of node Q.
However, in the simulation for structures that have curved surface, these two displacements might be
different from each other. In such case, it is desirable to use the same nodes for measuring crack opening
displacement and measuring nodal force [27]. We called this method two-step VCCT approach. In this method,
finite element analyses are performed twice: before crack opening and after crack opening, and nodal force and
relative displacement are calculated at the same node. In this method, it is unnecessary to assume similarity of
crack, because energy release rate is calculated from two conditions. While computational cost of this method is
simply twice as much as that of standard VCCT. In this study, we used standard VCCT in optimizations and
parametric studies because we have to perform a lot of simulations for optimizations. In this study, we used
standard VCCT in optimizations and parametric studies owing to computational cost, and then we compered the
results of standard VCCT and that of two-step VCCT in order to confirm the validity of the results of standard
VCCT.

3. Configuration of CFRP cryogenic tank

In this study, we conducted two types of optimizations: optimization of liner thickness and that of
curved shape of dome part. It is assumed that the tank is used at -196 C and maximum expected operating
pressure (MEOP) is 5 MPa. We set the diameter of cylinder part to 1 m and the diameter of cap to 0.2 m.
Aluminum alloy (A6061) is utilized as liner, and it is wrapped by CFRP. We set the thickness of CFRP so that
burst pressure became 10 MPa. We considered the turnaround of prepreg tape in the cap region. It was assumed
that CFRP was wound by only helical winding at the dome part, while at the cylindrical part, CFRP was
composed by two layers: helical and hoop layers. Figure 4 shows the schematic view of tank. The purpose of
optimization is to obtain the shape to prevent propagation of crack near the cap. Therefore FEM analyses were
merely conducted for the dome part. However, in order to consider the effect of the cylinder part, short cylinder
was attached at the root of the dome. FEM analyses were conducted in two conditions. In one condition, the tank
was cooled down to -196 C from 120 C which is the temperature of CFRP/Al liner bonding, and the
temperature was regarded as stress free temperature. In this study, the value of object function of this condition
is labelled as Cooling. In another condition, we applied the internal pressure of 5 MPa which is equal to
MEOP to the tank model in after cooling. The value of object function of this condition is labelled as Pressure.
We conducted each type of optimizations: liner thickness and curved shape of the dome part, under each
condition: Cooling and Pressure. Therefore, we conducted four optimizations in total. In FEM analyses, we
utilized axisymmetric solid elements, CGAX4R and CGAX3 in Abaqus/Standard. At helical layer, we defined
the winding angle of fiber based on geodesic curve as shown in equation (5).

sin = ro r (5)
where ro and r represent the radius of cap and the dome portion at a certain position, respectively.
We set the fracture criterion as shown in equation (6).
m m
G G
f = I + II
G


(6)
G IC IIC
where GI and GII represent mode I and mode II energy release rate, and GIC and GIIC represent mode I and mode
II fracture toughness respectively. We employed the power law as fracture criterion. We decided the multiplier
of criterion, GIC and GIIC based on MMB, DCB, and ENF tests at cryogenic temperature [26, 28-30] as shown in
appendix. We used GI C = 1.3 kJ/m2GIIC = 1.0 kJ/m2 , and m = 1. If f becomes larger than 1, crack propagation
occurs. We conducted standard VCCT analysis for the structure that configuration is described above. The
structure, which is called isotensoid, was designed by the netting theory. Table 1 and table 2 show the material
properties of CFRP and Al alloy. In this analysis, we set crack length to 5 mm from the cap. Figure 5 shows the
result of this analysis. From this analysis, the values of fracture criterion are more than 1 and it indicates crack
propagate in both conditions.

Table 1 Material properties of unidirectional CFRP

Temperature 23 C -196 C
Longitudinal modulus E1 159.6 GPa 161.3 GPa
Transverse modulus E2 (= E3) 8.73 GPa 12.67 GPa
Longitudinal shear modulus G12 (= G13) 4.51 GPa 7.18 GPa
Transverse shear modulus G23 3.01 GPa 4.37 GPa
Major Poissons ration 12 (= 13) 0.328 0.328
Transverse Poissons ratio 23 0.45 0.45
-9
Longitudinal coefficient of thermal expansion 1 -6.6210 1/K -3.1810-8 1/K
Transverse coefficient of thermal expansion 2 (=3) 3.7010-5 1/K 3.1010-5 1/K

Table 2 Material properties of aluminum alloy

Temperature 23 C -196 C
Young modulus E 70.9 GPa 77.1 GPa
Poisson ratio 0.33 0.35
-5
Coefficient of thermal expansion 2.1010 1/K 1.6810-5 1/K

4. Problem setting of optimizations and results

4.1 Method of optimization
We employed genetic algorithm, which performs probabilistic search, for optimizations. In first
generation, individuals were generated in a range of design values randomly. We used the elite preservation
strategy, which is the method to select superior individuals for next generation. We stopped optimizations when
the number of generation became 200. Flowchart of the optimization is shown in Fig. 6.
 4.2 Optimizations of liner thickness
We conducted optimization about liner thickness. We used isotensoid in curved shape and set the design
variables as shown in Fig.7. Thickness of each point which is indicated by dotted line in Fig. 7 was design
variable. Design variables xi were placed every 10 mm in the region where distance from the cap is smaller than
60 mm. We also placed a design variable at the point where the distance from the cap is 93 mm in order to
connect smoothly with the other region. Thickness of the region where the distance from the cap is larger than 93
mm was set to 2 mm. The upper limit of the design variables was set to 9 mm, because the height of the cap was
18 mm. We used the half value of the cap height for the upper limit. The lower limit of the design variables was
set to 2 mm due to limit of the manufacture. Objective function and the range of design variables are shown in
equation (7), and we used equation (8) for calculation of fitness.

GI G II
minimize f ( x ) = G + G
IC IIC

2 x i 9 (7)
x = 1
i

fitness = 1 / (1 + f (x)) (8)
,where x is vector of design variables and xi is incremental quantity of design variables.
The results of liner thickness optimization are shown in Figs.8 and 9. In Fig 8, the shapes after
optimizations are shown. Figure 9 shows comparison of the values of objective function of initial shape and
optimized shape. In the optimized shapes, thickness of the liner around the crack tip became very small. In such
shapes, this notch-like shape partially reduces stiffness of the liner and prevents opening of crack. Optimal
shapes for both cooling and pressure have similar notch-like shape and both of them can reduce the value of
objective function. Then, the structure that has the notch is effective for prevent propagation of crack. However
in Fig.9, because the value of initial structure is more than 1.0, crack progress and the tank may be broken. It
should be noted that the same concepts are seen in Ref. [19, 31]. In these concepts, structures have the part of
low stiffness. This part can relax the thermal stress.

4.3 Optimization for dome shape

We conducted optimization about curved surface shape. The thickness of the liner near the cap was
fixed in this optimization. We defined the curved surface shape as shown in Fig. 10. We used isotensoid to
define the initial value Ri. Angles i in Fig. 10 were fixed, and parameters ri were subtracted from Ri. Curved
surface shape Ri was calculated as Ri -ri . Design points, which constitute the dorm part, were 44 points in this
optimization. We used 20 parameters r near the cap and position of the cap d as design variables (see Fig. 10).
Therefore there were 21 design variables in this optimization. The upper and lower limits were set due to
consideration of computational cost. Object function and range of design variables are shown in equation (9),
and we used equation (8) to calculate fitness.
Figure 11 shows the portion which strongly affected the value of objective function in curved surface
shape optimization. In the optimal shape for cooling, flexure deformation which was caused by thermal
shrinkage was suppressed near the cap during cooling. On the other hand, the optimal shape for pressure
reduces the value of object function by the deformation due to the pressure. Because of pressure, bending
deformation occurs in this shape, and it closes the crack. Figure 12 shows the comparison of the value of object
function of initial shape and optimal shapes.

GI G II
minimize f (r , d ) = G + G
IC IIC
0.63 ri 0.64

ri = 0.01 (9)
6.3 d 6.4

d = 0.1


,where r is vector of design variables to modify dome shape and d is a design variable to change the
position of the cap. ri and d are incremental quantities of design variables.

5. Parametric studies and verification of VCCT

From the two optimizations described above, the following results were obtained.

Liner thickness effects on energy release rate and notch structure was generated from optimization. Notch
can reduce the objective function f.

From dome shape optimizations, optimal shape for both of cooling and pressure condition can reduce
fracture criterion. Especially, optimal shape for pressure condition can prevent crack opening because
deformation of pressure condition closes crack in this structure. Thus using the shape like Fig.11 (b) can
reduce the objective function f.
However, the optimal position of notch is not clear due to setting of design variable. Therefore we conducted
parametric study in order to decide precise position of the notch. In this parametric study, we use the shape that
was generated in optimization (in Fig.11 (b)) to consider the deformation of pressure, because it is severer than
that of cooling. Then we changed the position of notch from 1 mm to 10 mm from the cap, and observed
tendency of the change of the objective function f. We set the depth of the notch to 5.5 mm, which was obtained
from the optimization, and the thickness of liner near the cap with 7.5 mm (see Figure 13).
Figure 14 shows the result of parametric study for position of notch. The value of objective function
decreased as the position of notch got closer to the cap when the distance from the cap was larger than 3.0 mm.
In contrast, the value of the objective function increased when the distance from the cap to the notch was less
than 3 mm. In the regions from 3 mm to 5 mm from the cap, the values of the objective function are less than 1.0
in condition of Pressure, and it can prevent propagation of crack. The value when the notch position was 4 mm
was under 1.0 in both of condition, Cooling and Pressure. Figure 15 shows the deformation shape, in which
the notch was placed at the position of 4 mm from the cap. In this figure, when the position of the notch is 4mm,
this structure can prevent deformation of crack opening. Figure 16 shows comparison of the values of the
objective function between initial shape and the proposed shape shown in Fig. 13. From this figure, fracture
criterion of proposed design is much lower than that of initial shape.
We compered the value of fracture criterion that is calculated from standard VCCT and two-step VCCT
in order to confirm the accuracy of this analysis. In two-step VCCT, weak spring and penalty spring were used
for calculation of relative displacement and nodal force. Figure 17 shows relationship between crack length and
the value of fracture criterion. These results indicate energy release rate obtained from standard VCCT tends to
overestimate slightly than two-step VCCT. Thus in optimizations and parametric studies, condition of analyses
in standard VCCT is harder than that of two-step VCCT. However, tendency of the values of standard VCCT are
good agreement with that of two-step VCCT in all crack length and standard VCCT is applicable to this study.
We can conclude that this design can be used under deformation of cooling (at -196 C) and pressure (5 MPa).

6. Conclusion
In this study, we conducted structural optimizations for CFRP cryogenic tank by using actual material
properties, which are obtained from DCB, ENF and MMB tests in cryogenic temperature. Energy release rate of
debonding crack calculated by virtual crack closure method was used as objective function and we used
thickness of liner and dome shape for design variables in optimizations. We conducted optimization for liner
thickness and dome shape in different conditions (cooling and pressure condition). From optimizations for liner
thickness, we obtained the notch-like shape in both cooling and pressure conditions. These notch-like shapes
reduce stiffness of liner and prevent crack propagation. From optimizations for dome shapes, the optimal shape
for Cooling suppressed flexure deformation caused by thermal shrinkage, and the optimal shape for Pressure
closes crack by bending deformation because of pressure. Finally, we proposed the structure that was combined
notch-like shape and optimal dome shape and conducted parametric studies about position of notch. From these
parametric studies, if we provide the notch in suitable position, it is possible to reduce energy release rate highly
and prevent propagation of crack.

Appendix
DCB, ENF and MMB tests at cryogenic temperature
In order to obtain mode I and mode II energy release rate, we conducted double cantilever bending
(DCB) test and end notched fracture (ENF) test at cryogenic temperature (123 K). Mixed mode bending (MMB)
test was also conducted for the purpose of determining fracture criterion. Adhered CFRP (Toray, T800/3900-2B)
was used for test specimens as shown in figure A1. Cure temperature of this CFRP is 180 C and stacking
sequence of specimens was [0]12. Epoxy resin adhesive film (3M, AF163-2K) was used for adhesive. We did
sanding and degreasing on surface of CFRP.
We conducted DCB tests to use JIS K 7086 as reference. For CFRP, we introduced pre crack whose
length was 23 mm (see figure A1). In order to apply load, displacement was given vertically for CFRP.
Experiments were conducted in thermostat. Therefore, we used crack gage (Kyowa, KV-25) for measuring crack
propagation. ENF tests were also conducted in same experimental conditions and JIS K 7086 was referred.
However length of pre crack was 28 mm and jig span was 100 mm due to jig. In testing, crack propagated
unstably. Therefore, crack length was not measured during experiments. Experimental conditions of MMB tests
were same as DCB and ENF tests and we referred to ASTM D 6671. However because of jig span, length of pre
crack 28 mm and jig span was 100 mm. In testing, we set the mixed-mode ration to R = 0.5. The results of three
tests are summarized in figure A2. According to figure A2, we decided the mixed fracture criterion by using
power law as,
m n
G G
G = I + II
G

(A1)
GIC IIC
where GI C = 1.3 kJ/m2GIIC = 1.0 kJ/m2 , and m= n = 1.
 References
1. Tecnical Notes. ARS Journal. 1962;32:943-82.
2. Fukunaga H, Uemura M. Optimum design of helically wound composite pressure vessels. Composite
Structures. 1983;1:31-49.
3. Vasiliev VV, Krikanov AA, Razin AF. New generation of filament-wound composite pressure vessels for
commercial applications. Composite Structures. 2003;62:449-59.
4. Zu L, Koussios S, Beukers A. Design of filamentwound domes based on continuum theory and non-geodesic
roving trajectories. Composites Part A: Applied Science and Manufacturing. 2010;41:1312-20.
5. Onder A, Sayman O, Dogan T, Tarakcioglu N. Burst failure load of composite pressure vessels. Composite
Structures. 2009;89:159-66.
6. Kim CU, Hong CS, Kim CG, Kim JY. Optimal design of filament wound type 3 tanks under internal pressure
using a modified genetic algorithm. Composite Structures. 2005;71:16-25.
7. Kim C-U, Kang J-H, Hong C-S, Kim C-G. Optimal design of filament wound structures under internal
pressure based on the semi-geodesic path algorithm. Composite Structures. 2005;67:443-52.
8. Parnas L, Katrc N. Design of fiber-reinforced composite pressure vessels under various loading conditions.
Composite Structures. 2002;58:83-95.
9. Holland JH. Building blocks, cohort genetic algorithms, and hyperplane-defined functions. Evol Comput.
2000;8:373-91.
10. Kirkpatrick S, Gelatt CD, Vecchi MP. Optimization by Simulated Annealing. Science. 1983;220:671-80.
11. Medzhitov R. Recognition of microorganisms and activation of the immune response. Nature. 2007;449:819-
26.
12. Xu P, Zheng JY, Liu PF. Finite element analysis of burst pressure of composite hydrogen storage vessels.
Materials & Design. 2009;30:2295-301.
13. Xu P, Zheng JY, Chen HG, Liu PF. Optimal design of high pressure hydrogen storage vessel using an
adaptive genetic algorithm. International Journal of Hydrogen Energy. 2010;35:2840-6.
14. Vafaeesefat A. Optimization of composite pressure vessels with metal liner by adaptive response surface
method. Journal of Mechanical Science and Technology. 2011;25:2811-6.
15. Cohen D. Influence of filament winding parameters on composite vessel quality and strength. Composites
Part a-Applied Science and Manufacturing. 1997;28:1035-47.
16. Chang RR. Experimental and theoretical analyses of first-ply failure of laminated composite pressure vessels.
Composite Structures. 2000;49:237-43.
17. Hwang TK, Hong CS, Kim CG. Size effect on the fiber strength of composite pressure vessels. Composite
Structures. 2003;59:489-98.
18. Hwang T-K, Hong C-S, Kim C-G. Probabilistic deformation and strength prediction for a filament wound
pressure vessel. Composites Part B: Engineering. 2003;34:481-97.
19. Takeuchi S, Sato E, Onoda J, Arakawa Y, Miyahara A. Contribution of Reduction Method for Energy Release
Rate of Adhesive Bonding Structure in Composite Tank around Mouthpiece. JOURNAL OF THE JAPAN
SOCIETY FOR AERONAUTICAL AND SPACE SCIENCES. 2010;58:302-7.
20. Liu PF, Chu JK, Hou SJ, Xu P, Zheng JY. Numerical simulation and optimal design for composite high-
pressure hydrogen storage vessel: A review. Renewable & Sustainable Energy Reviews. 2012;16:1817-27.
21. NASA, Final Report of the X-33 Liquid Hydrogen Tank Test Investigation Team. 2000.
22. Morimoto T, Ishikawa T, Yokozeki T, Hayashi Y, Shimoda T, Morino Y. Pressurization test on CFRP liner-
less tanks at liquefied nitrogen temperature. Advanced Composite Materials. 2004;13:81-8.
23. Schneider Judy DM, Hastings Chad, Patterson James, Noorda Jared, DeLay Thomas. Lightweight Cryogenic
Composit Over-wrapped Pressure Vessels (COPVS) for Launch Vehicle Applications. 48th
AIAA/ASME/ASCE/AHS/ASC Structures. Structural Dynamics, and Materials Conference. 2007.
24. Ju J, Pickle BD, Morgan RJ, Reddy JN. An Initial and Progressive Failure Analysis for Cryogenic Composite
Fuel Tank Design. Journal of Composite Materials. 2007;41:2545-68.
25. Tomassetti G, Barboni R, de Benedetti M. Optimisation methodology for cryotanks. Computers & Structures.
2005;83:2293-305.
26. Yoshimura A, Takaki T, Noji Y, Yokozeki T, Ogasawara T, Ogihara S. Fracture Toughness of CFRP Adhesive
Bonded Joints at Cryogenic Temperature. Journal of Adhesion Science and Technology. 2012;26:1017-31.
27. Krueger R. Virtual crack closure technique: History, approach, and applications. Applied Mechanics Reviews.
2004;57:109.
28. Aoki T, Ishikawa T, Kumazawa H, Morino Y. Cryogenic mechanical properties of CF/polymer composites
for tanks of reusable rockets. Advanced Composite Materials. 2001;10:349-56.
29. Nairn JA. Energy release rate analysis for adhesive and laminate double cantilever beam specimens
emphasizing the effect of residual stresses. International Journal of Adhesion and Adhesives. 2000;20:59-70.
30. Yokozeki T, Ogasawara T, Aoki T, Ishikawa T. Experimental evaluation of gas permeability through damaged
composite laminates for cryogenic tank. Composites Science and Technology. 2009;69:1334-40.
31. Akinori Y, Toshio O, Hiroshi S. Validation of thermal stress reduction design of CFRP cryogenic propellant
tank by LN2 immersion test. JAXA research and development memorandum. 2013;JAXA-RM-12-012.
Figure captions
Fig. 1 Distribution of stress and displacement around crack tip.
Fig. 2 Nodal force and relative displacement in elements.
Fig. 3 Schematic of two-step VCCT.
Fig. 4 Schematic view of CFRP cryogenic tank.
Fig. 5 Initial values of fracture criterion f, which were calculated from the structure of isotensoid.
Fig. 6 Flowchart of method of optimization, where FEM model is created by Abaqus / CAE and
VCCT analysis is carried out in Abaqus / standard.
Fig. 7 Optimization points of liner thickness.
Fig. 8 Solution obtained by the optimization of liner thickness.
Fig. 9 Comparison of energy release rate with initial model and optimized liner thickness obtained by liner
thickness optimization.
Fig. 10 Schematic view of change in the dome shape where the dashed line represents isotensoid and the solid
line represents curved surface obtained by GA.
Fig. 11 Solution obtained by the optimization of curved surface shapes.
Fig. 12 Comparison of energy release rate with initial model and optimized model obtained by curved surface
shapes optimization.
Fig. 13 Schematic view for the location of the notch.
Fig. 14 Relationship of fracture criterion f and notch position.
Fig. 15 Shape obtained by the parameter study.
Fig. 16 Comparison of fracture criterion of initial shape and proposed design.
Fig. 17 Comparison of fracture criterion between standard VCCT and two-step VCCT.
Fig. A1 Specimen appearance of DCB, ENF and MMB tests in cryogenic temperature.
Fig. A2 Relationship between mode I fracture toughness and mode II fracture toughness at cryogenic
temperature (123K).
Figure 1 Distribution of stress and displacement around crack tip.
 2

f1
Q
1
P f2

a a
Figure 2 Nodal force and relative displacement in elements.
 2
Displacement Nodal force

f1
1
f2
Weak spring
Penalty spring
a a
Figure 3 Schematic of two-step VCCT.
Figure 4 Schematic view of CFRP cryogenic tank.
 5

4
Fracture criterion f

0
Cooling Pressure
Figure 5 Initial values of fracture criterion f, which were calculated from the structure of isotensoid.
 Creating initial population
by using GA

Generation n = 1

Creating FEM model

by Abaqus / CAE

VCCT analysis
by Abaqus / standard

Calculation of fitness

Selection
by using roulette selection

Cross-over and mutation

n++

Max generation No
<n

End

Figure 6 Flowchart of method of optimization, where FEM model is created by Abaqus / CAE and
VCCT analysis is carried out in Abaqus / standard.
Figure 7 Optimization points of liner thickness.
 Crack

Aluminum alloy CFRP

liner

Notch
(a) Cooling optimization

Crack

CFRP
Aluminum alloy
liner

Notch
(b) Pressure optimization
Figure 8 Solution obtained by the optimization of liner thickness.
 5
4.5 Cooling
Pressure
4
Fracture criterion f

3.5
3
2.5
2
1.5
1
0.5
0
Initial Cooling Optimization Pressure Optimization
Figure 9 Comparison of energy release rate with initial model and optimized liner thickness obtained by
liner thickness optimization.
 Isotensoid

d
ri+2 Modified Surface

Ri+2 Ri+1
Ri

i i+1
Figure 10 Schematic view of change in the dome shape where the dashed line represents isotensoid and the
solid line represents curved surface obtained by GA.
 Crack
CFRP
Aluminum alloy
liner

(a) Cooling optimization

Crack

Aluminum alloy
liner
CFRP

(b) Pressure optimization

Figure 11 Solution obtained by the optimization of curved surface shapes.
 5
4.5 Cooling
Fracture criterion f 4 Pressure

3.5
3
2.5
2
1.5
1
0.5
0
Initial Cooling Optimization Pressure Optimization
Figure 12 Comparison of energy release rate with initial model and optimized model obtained by curved
surface shapes optimization.
 Notch

CFRP

Liner
Figure 13 Schematic view for the location of the notch.
 10 Cooling
Fracture criterion f Pressure
8

0
1 2 3 4 5 6 7 8 9 10
Notch position [mm]
Figure 14 Relationship of fracture criterion f and notch position.
Mode I energy release rate Pressure deformation
0.0 3.0 (kJ/m2) Deformation magnitude = 5

CFRP CFRP

Liner Liner

(a) 4 mm (b) 10 mm
Figure 15 Shape obtained by the parameter study.
 5
Cooling
Fracture criterion f
4 Pressure

0
Initial Proposed design
Figure 16 Comparison of fracture criterion of initial shape and proposed design.
 1.5
Standard VCCT (Cooling)
Standard VCCT (Pressure)
Fracture craiterion f
Two-step VCCT (Cooling)
1 Two-step VCCT(Pressure)

0.5

0
1 2 3 4 5
Crack length [mm]
Figure 17 Comparison of fracture criterion between standard VCCT and two-step VCCT.
 Film Adhesive CFRP
12.5 mm Thickness
CFRP : 2.31 mm
Adhesive : 0.2mm
CFRP

23 mm 27mm 100 mm
(a) DCB, MMB specimen

Film Adhesive CFRP

12.5 mm

CFRP

50 mm 100 mm
(b) ENF specimen

Figure A1 Specimen appearance of DCB, ENF and MMB tests in cryogenic temperature.
 Mode I fracture toughness GI [kJ/m ]
2 2

1.5

0.5

0
0 0.5 1 1.5 2
2
Mode II fracture toughness GII [kJ/m ]
Figure A2 Relationship between mode I fracture toughness and mode II fracture toughness at cryogenic
temperature (123K).