Вы находитесь на странице: 1из 14

Desalination 397 (2016) 126–139

Contents lists available at ScienceDirect

Desalination

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

Analytical and finite element modeling of pressure vessels for seawater


reverse osmosis desalination plants
A.M. Kamal, T.A. El-Sayed ⁎, A.M.A. El-Butch, S.H. Farghaly
Department of Mechanical Design, Faculty of Engineering, Helwan University, P.O. Box 11718, Mataria, Cairo, Egypt

H I G H L I G H T S

• The pressure vessel for SWRO has been modeled using analytical solution and finite element modeling.
• Comparison between stainless steel and fiber reinforced composite materials has been done.
• The design parameters of pressure vessel have been optimized.
• The results of both analytical solution and finite element modeling have showed good agreement.

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

Article history: A pressure vessel (PV) which contains the membrane elements of seawater reverse osmosis (SWRO) desalina-
Received 24 October 2015 tion has been modeled using analytical solution and finite element modeling (FEM) to optimize the PV design
Received in revised form 14 June 2016 parameters. Two types of PV materials have been compared namely; stainless steel and fiber reinforced compos-
Accepted 16 June 2016
ite materials. Von-Mises yield criterion and Tsai-Wu failure criterion are used for the design of stainless steel and
Available online 5 July 2016
composite PVs respectively. E-glass/epoxy and carbon/epoxy composite materials are considered in this work. In
Keywords:
addition, hybrid composite materials are introduced for layers through the vessel thickness. The results have
Pressure vessel shown that the optimum lay-up is achieved using the angle-ply [± ϴ]ns at winding angle of 54° for E-glass/
Sea water desalination plant epoxy and 55° for carbon/epoxy PVs while for hybrid composite PVs the optimum lay-up is [90G/±50C/90G]ns.
Composite materials Also, the results have shown that the composite PVs have lighter weight than the stainless steel PVs. The
Analytical and FEM carbon/epoxy PVs introduce the optimum weight savings but in terms of the total PVs cost, the hybrid composite
Membrane unit PVs can be used.
© 2016 Elsevier B.V. All rights reserved.

Nomenclature (continued)

Symbol Meaning Units


Symbol Meaning Units
Ftu Tensile ultimate strength MPa
a, b, c, d Laminate compliance matrices MPa−1 Fty Tensile yield strength MPa
A, B, D Laminate stiffness matrices MPa F1c Longitudinal compressive strength MPa
Ai Flow area mm2 F1t Longitudinal tensile strength MPa
cθ Cosine of winding angle – F2c Transverse compressive strength MPa
Di Inner diameter of PV mm F2t Transverse tensile strength MPa
E Young's tensile modulus MPa F6 In-plane shear strength MPa
E1 Longitudinal tensile modulus MPa G12 Shear modulus in the 1–2 plane MPa
E2 Transverse tensile modulus (in 2-direction) MPa G13 Shear modulus in the 1–3 plane MPa
E3 Transverse tensile modulus (in 3-direction) MPa G23 Shear modulus in the 2–3 plane MPa
f1, f2 Tsai-Wu coefficients MPa−1 h Thickness of pressure vessel mm
f11, f12, f22, f66 Tsai-Wu coefficients MPa−2 hk Coordinate of layer k upper surface mm
Fa Side thrust N hk-1 Coordinate of layer k lower surface mm
i Number of layers through vessel thickness –
k Layer order –
Abbreviations: PV, pressure vessel; RO, reverse osmosis; SWRO, seawater reverse L Length of PV mm
osmosis; FEM, finite element modeling. Mx,y Moments matrix per unit length N
⁎ Corresponding author. Mx, My, Ms Bending and twisting moments per unit length N
E-mail address: TAMER_ALSAYED@m-eng.helwan.edu.eg (T.A. El-Sayed). n Number of basic laminates –

http://dx.doi.org/10.1016/j.desal.2016.06.015
0011-9164/© 2016 Elsevier B.V. All rights reserved.
A.M. Kamal et al. / Desalination 397 (2016) 126–139 127

(continued) 1. Introduction
Symbol Meaning Units
Reverse osmosis (RO) is considered one of the most efficient
nc Number of carbon/epoxy layers –
ng Number of E-glass/epoxy layers –
methods for seawater desalination and many researches have been de-
Nx,y Forces matrix per unit length N/mm voted to optimize this process. The main components of a seawater re-
Nx, Ny, Ns Axial, hoop and shear forces per unit length N/mm verse osmosis (SWRO) desalination plant are seawater intake,
p Internal fluid pressure MPa pretreatment, reverse osmosis system and post-treatment is shown in
Qx,y Transformed lamina stiffness matrix w. r. t. global axes MPa
Fig. 1 [1]. Spiral wound membrane elements (from 1 to 6 elements)
Q1,2 Lamina stiffness matrix w. r. t. material axes MPa
Q11, Q12, Q22, Components of Q1,2 MPa are installed in a PV which is usually fabricated from stainless steel or
Q66 fiber-reinforced composite materials [2], see Fig. 2.
sθ Sine of winding angle – Much research have been developed to optimize the design param-
Sf Safety factor of PV – eters of composite PVs such as properties of fiber and matrix, fiber
Sfmin Minimum safety factor –
Sfk Safety factor of layer k –
winding angle, fiber concentration and number of layers through the
ΔSf Safety factor error – vessel thickness.
t Ply (layer) thickness mm Krikanov [3] proposed a method to design hybrid laminated com-
T Transformation matrix – posite PVs under different loading conditions. The vessel consists of a
Vf Fiber volume fraction –
hoop layer reinforcing the cylindrical part and a helical layer reinforcing
Wcomp Weight of composite PV kg
Wst Weight of stainless steel PV kg the vessel domes. He concluded that when the vessel axial strain is
z Coordinate of layer midplane w. r. t. reference plane mm constrained, transformation between the helical layer and the hoop
α Coefficient of S2fk – layer should be done with a suitable angle and when the hoop strain
β Coefficient of Sfk – is constrained, the hoop layer material should be replaced with higher
γs In-plane shear strain w. r. t. global axes –
stiffness material. Xu et al. [4] proposed a 3D parametric finite element
γos In-plane shear strain of reference plane –
γ4 Shear strain in the 2–3 plane model to estimate the failure evolution and strength of composite hy-
γ5 Shear strain in the 1–3 plane – drogen storage vessels. They established a solution algorithm using
γ6 Shear strain in the 1–2 plane – ANSYS finite element software to examine the damage progress and
ϵ Normal strain –
the failure properties of composite structures with increasing internal
ϵx Axial strain (in x-direction) –
ϵy Hoop strain (in y-direction) –
pressure using different failure criteria. Their results show a good agree-
ϵx, y Strains matrix w. r. t. global axes – ment between the theoretical failure pressure and the experimental
ϵo x;y Reference plane strains matrix – burst pressure especially for Tsai-Wu failure criterion.
ϵox, ϵoy Axial and hoop strains of reference plane – Son and Chang [5] calculated the stress distributions in the compos-
ϵu1t Ultimate longitudinal tensile strain –
ite layers of a hydrogen PV by using three different modeling techniques
ϵu2t Ultimate transverse tensile strain –
ϵ1 Longitudinal tensile strain (in 1-direction) – which are laminate-based, full ply-based and hybrid (combining a lam-
ϵ2 Transverse tensile strain (in 2-direction) – inate-based modeling for the dome part with a ply-based modeling for
ϵ1,2 Strains matrix w. r. t. material axes – the cylinder part). The models were created using ABAQUS 6.9-1com-
ϵ3 Transverse tensile strain (in 3-direction) – mercial finite element software. The results showed that the PV failure
ϴ Fiber orientation (winding angle) Deg.
κx, y Reference plane curvatures matrix –
occurred by the transverse tensile stress at the border under the test
κx, κy, κs Components of κx,y – pressure of (105 MPa) which is 1.5 of the operating pressure
ρc Density of carbon/epoxy material kg/m3 (70 MPa). The finite element results were verified using experimental
ρcomp Density of composite material kg/m3 results of a PV prototype. The results showed that complex geometries
ρg Density of E-glass/epoxy material kg/m3
can be simply modeled using the laminate-based modeling technique
ρst Density of stainless steel kg/m3
ν Poisson's ratio – and a reasonable overall stress distribution can be obtained. Moreover,
ν12 Poisson's ratio at load in 1-direction and strain in – the full ply-based modeling technique can be used to obtain more accu-
2-direction rate stress distributions results in both the metal liner and the compos-
ν13 Poisson's ratio at load in 1-direction and strain in – ite layers.
3-direction
ν21 Poisson's ratio at load in 2-direction and strain in –
Onder et al. [6] studied the effect of winding angle of composite PVs
1-direction on the burst pressure using an analytical solution and finite element
ν23 Poisson's ratio at load in 2-direction and strain in – method. They performed an experimental method to verify the opti-
3-direction mum winding angle for symmetric and antisymmetric shells. The ana-
ν31 Poisson's ratio at load in 3-direction and strain in –
lytical and experimental results were in good agreement for some
1-direction
ν32 Poisson's ratio at load in 3-direction and strain in – orientations.
2-direction Parnas and Katirci [7] used the classical lamination theory and gen-
σ Normal stress MPa eralized plane strain model to derive an analytical solution to evaluate
σmax Maximum normal stress MPa the behavior of fiber reinforced composite PVs. This solution includes
σmin Minimum normal stress MPa
σx Axial stress (in x-direction) MPa
the effect of the internal pressure, axial force and body force due to ro-
σy Hoop stress (in y-direction) MPa tation. In addition, the effect of temperature and moisture variation is
σz Radial stress (in z-direction) MPa considered. Xie et al. [8] introduced a technique for the optimization
σx,y Stresses matrix w. r. t. global axes MPa of fiber orientations and weighting factors in hybrid-fiber multilayer-
σ1 Longitudinal tensile stress (in 1-direction) MPa
sandwich cylindrical shells subjected to external pressure. The results
σ2 Transverse tensile stress (in 2-direction) MPa
σ1 ,2 Stresses matrix w. r. t. material axes MPa of the examples presented in their work showed that the maximum
σ3 Transverse tensile stress (in 3-direction) MPa critical pressure can be obtained when using higher modulus fiber in
τ Shear stress MPa the hoop direction and lower modulus fiber in the longitudinal
τmax Maximum shear stress MPa direction.
τs Shear stress in the x-y plane MPa
τ4 Shear stress in the 2–3 plane MPa
Mian et al. [9] developed an optimization method for composite PV
τ5 Shear stress in the 1–3 plane MPa lay-ups using finite element analysis and calculated the relative weight
τ6 Shear stress in the 1–2 plane MPa saving compared with the reference aluminum PV. Three composite
materials namely; S-glass/epoxy, Kevlar/epoxy and carbon/epoxy are
128 A.M. Kamal et al. / Desalination 397 (2016) 126–139

Fig. 1. Schematic diagram of a typical SWRO desalination plant [1].

used. They also established a MATLAB code to verify their results. They uniform stress distribution between the steel liner and the composite
concluded that the optimum lay-up is achieved at the angle-ply layers is obtained when using carbon fibers materials. Daniel and Ishai
[± ϴ]ns and at a winding angle of 54° for all composite materials. [14] presented an example of a PV subjected to internal pressure and
Wang et al. [10] studied experimentally the properties of PVs made of external torque. They compared three types of composite materials
carbon and aramid fibers hybrid composites with epoxy resin at differ- with different lay-ups using an analytical approach. Then, they deter-
ent ply types and hybrid ratios. They concluded that the vessels with mined the optimum lay-up for each material which achieves the mini-
longitudinally unique fiber type have high strength and positive hybrid mum vessel thickness. Finally, they calculated the relative weight
effect while the vessels with longitudinally different fiber types have savings for the optimum PV compared with an aluminum reference PV.
low strength and negative hybrid effect. The present research focused on optimization of the SWRO (PV) de-
Assam et al. [11] studied the influence of several design parameters sign parameters using both analytical solution and FEM. Comparison
on the performance of filament-wound composite PVs. These PVs con- between stainless steel and composite material PVs is evaluated with
sists of inner liner and composite layers. The material properties and respect to material saving, specific strength and cost. Four different
the thickness of the pressure vessel components are investigated. sizes of PVs are used which are the standard sizes available in the inter-
Their results show that for large number of layers the effect of liner on national markets. For composite material PVs, E-glass/epoxy, carbon/
the internal pressure is very limited. In addition, they showed that in- epoxy and hybrid materials are selected with five symmetric laminate
creasing both the number of layers and the layer thickness results in lay-ups.
gain in the operating efficiency and the ultimate failure pressure.
Taghavian et al. [12] presented a new method for designing compos- 2. Analytical solution
ite PVs under different loading and constraints conditions using lattice
structures. They developed an analytical approach and verified the re- A thin-walled cylindrical PV subjected to internal pressure is investi-
sults experimentally. They concluded that significant material saving gated using linear elasticity analysis. Two types of PV materials have
can be achieved by using of lattice structures for hoop strain suppress- been compared namely; stainless steel and fiber reinforced composite
ing compared with that of addition of extra plies while for axial strain materials, to optimize the PV design parameters. The optimization pro-
suppressing, no considerable material saving is recognized. cess is carried out by using a computational program established by
M. Sabour and Foghani [13] proposed a method to design a metal- MATLAB (R2012b).
composite PV using a combination of finite element method and fuzzy
decision making. They used three types of fibers which are carbon, 2.1. Analysis of stainless steel PV
glass and Kevlar and optimized the vessel weight for each type. The re-
sults showed that PVs with carbon fibers have minimum weight and The SWRO stainless steel PV is considered as an open pipe subjected
high strength compared with other PVs. They also concluded that a to internal pressure and side thrust exerted at both sides of the pipe due

Fig. 2. Spiral wound element PV assembly [2].


A.M. Kamal et al. / Desalination 397 (2016) 126–139 129

to the effect of end caps on the vessel, see Fig. 3. The value of axial forces, Table 1
Fa in N, is calculated, neglecting the areas of feed, reject and product Properties of stainless steel 316L [16].

ports, as follows: Properties Values

Density, ρst, kg/m3 7750


F a ¼ pAi ð1Þ
Young's modulus, E, GPa 193
Poisson's ratio, ν 0.31
π 2 Tensile yield strength, Fty, MPa 207
; Ai ¼ D ð2Þ
4 i Tensile ultimate strength, Ftu, MPa 586

where,

Using the von-Mises yield criterion for the design of stainless steel
p is the internal fluid pressure in MPa
PV as follows:
Di is the inner diameter of PV in mm
Ai is the flow area in mm2.  2
Four industrial sizes of PVs are selected to be 2.5, 4, 8 and 16 in. re- F ty
σ 2max þ σ 2min −σ max σ min ¼ ð9Þ
spectively according to the available sizes in the international markets. Sfall
The vessel length, L, is 0.4 m for 2.5 in. diameter and 1 m for the other
sizes which determined to be suitable for one membrane element. The Where, Sfall is the allowable safety factor (Sfall = 2).
value of internal pressure is selected to be 8 MPa (80 bar) which suitable From the above analysis, the stainless steel PV thickness can be cal-
for most SWRO desalination processes. culated as follows:
The longitudinal and hoop stresses, σx and σy in MPa, acting on an
pffiffiffi
element of the cylindrical shell along the axial and hoop directions (x 3 pDi
h¼ S ð10Þ
and y), see Fig. 3, are obtained as follows: 4 F ty fall

pDi
σx ¼ ð3Þ Then, the weight of the stainless steel PV, can be obtained as follows:
4h

π h i
pDi W st ¼
2
ρst ðDi þ 2hÞ −D2i L ð11Þ
σy ¼ ð4Þ 4
2h

where, h is the thickness of PV in mm. where Wst is the weight in kg and ρst is density in kg/m3.
For thin-walled PV, the radial stresses can be neglected, σz = 0,
(plane stress condition). 2.2. Analysis of composite PV
The principal stresses for this state of stress are
The SWRO multi-layered filament wound composite PV for the tube
σx þ σy
σ max ¼ þ τ max ð5Þ of membrane unit is analyzed at the same loading conditions (internal
2
pressure and side thrust) of the stainless steel PV, Fig. 3. The same inter-
σx þ σy nal diameters and lengths are used. Two types of composite materials
σ min ¼ −τ max ð6Þ are used namely; E-glass/epoxy and carbon/epoxy (AS4/3501-6) and
2
their properties are shown in Table 2 [14].
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
σ −σ 2 The PV wall thickness is determined for each vessel size at different
x y
τ max ¼ þ τ2 ; τ ¼ 0 ð7Þ types of symmetric laminate lay-ups, see Table 3, with the variation of
2
fiber orientation or winding angle (angle between fiber direction and
The stainless steel material is considered to be isotropic and linear axis of the vessel, ϴ in degree, see Fig. 4). Hybrid composite materials
elastic with the following stress-strain relation (Hook's law): are also used with the same lay-ups by introducing layers with different
composite materials as shown in Table 3. For example the lay-up
σ ¼ Eϵ ð8Þ [+ϴC/−ϴG]ns means that one layer of carbon/epoxy material at posi-
tive winding angle and the other layer of E-glass/epoxy material at neg-
Many stainless steel grades have been used for SWRO (PV) such as ative winding angle. The subscripts can be defined as follows:
austenitic grade 316L, highly alloyed austenitic grade 904L, duplex
grade 2205 and super duplex 2507 [15]. For cost considerations, the se-
Table 2
lected stainless steel grade in this work is 316L with the properties at
Properties of composite materials [14].
room temperature shown in Table 1 [16].
Carbon/epoxy
Properties E-glass/epoxy (AS4/3501-6)

Fiber volume fraction, Vf 0.55 0.63


Density, ρ, kg/m3 2100 1580
Longitudinal modulus, E1, GPa 39 142
Transverse modulus, E2, GPa 8.6 10.3
In-plane shear modulus, G12, GPa 3.8 7.2
Major Poisson's ratio, ν12 0.28 0.27
Minor Poisson's ratio, ν21 0.06 0.02
Longitudinal tensile strength, F1t, MPa 1080 2280
Transverse tensile strength, F2t, MPa 39 57
In-plane shear strength, F6, MPa 89 71
Ultimate longitudinal tensile strain, ϵu1t 0.028 0.015
Ultimate transverse tensile strain, ϵu2t 0.05 0.006
Longitudinal compressive strength, F1c, MPa 620 1440
Transverse compressive strength, F2c, MPa 128 228
Fig. 3. PV loading.
130 A.M. Kamal et al. / Desalination 397 (2016) 126–139

Table 3
Types of symmetric laminate lay-ups.

Lay-up [±ϴ]ns [0/±ϴ]ns [90/±ϴ]ns [0/±ϴ/90]ns [90/±ϴ/90]ns

Material E-glass/epoxy E-glass/epoxy E-glass/epoxy E-glass/epoxy E-glass/epoxy


Carbon/epoxy Carbon/epoxy Carbon/epoxy Carbon/epoxy Carbon/epoxy
[+ϴC/−ϴG]ns [0C/±ϴG]ns [90C/±ϴG]ns [0C/±ϴG/90C]ns [90C/±ϴG/90C]ns
[±ϴC/±ϴG]ns [0G/+ϴG/−ϴC]ns [90G/+ϴG/−ϴC]ns [0G/±ϴC/90G]ns [90G/±ϴC/90G]ns

n is the multiple of plies or number of basic laminates and s is the Where,


symmetric stacking sequence.
The analysis procedure of multidirectional laminates is considered σ3 = τ4 = τ5 =0, γ4 =γ5 =0, ϵ ≠ 0 (eliminated from Eq. (14))
using the classical lamination theory with its assumptions [14,17] and 8 9 2 38 9
applied for the basic laminate unit for each lay-up (n = 1). < σ1 = Q 11 Q 12 0 < ϵ1 =
or σ 2 ¼ 4 Q 12 Q 22 0 5 ϵ2 or in shortened form σ1;2 ¼ Q 1;2 ϵ 1;2 ð15Þ
The unidirectional composite material of each lamina is considered : ; : ;
τ6 0 0 Q 66 γ6
to be orthotropic and linear elastic with the following stress-strain rela-
tions (referred to the principal material axes 1, 2, 3) [14]: And the components of lamina stiffness Q1 , 2 can be determined as
2 3 follows:
1 ν 21 ν31
6 E − − 0 0 0 7
6 1 E2 E3 7 E1 ν 21 E1 E2
6 7
8 9 6 − ν 12 1

ν32
0 0 0 78 9 Q 11 ¼ ;Q ¼ ;Q ¼ ; Q ¼ G12 ð16Þ
1−ν12 ν 21 12 1−ν 12 ν 21 22 1−ν 12 ν 21 66
> ϵ1 > 6 E1 E2 E3 7> σ 1 >
>
> > 6 >
7> >
>
>
> ϵ2 >
>
>
>
6 ν 13
6 ν 23 1 7>
> σ2 >
>
>
>
>
< = 6− − 0 0 0 7
7< σ 3 = The transformed lamina stiffness Qx , y referred to global axes (x, y)
ϵ3 E1 E2 E3
¼6
6
7
7> τ4 > ð12Þ
>
> γ4 >
> 6 0 1 > > can be determines as follows:
>
> > 0 0 0 0 7> τ5 >
7>
>
> γ >> 6 G23 7>
>
>
: 5> ; 66
>
7: τ6 ;
> 8 9 8 9
γ6 6 0 1 < σ1 = < σx =
6 0 0 0 0 7
7
6 G13 7 σ ¼ T σy or in shortened form σ1;2 ¼ Tσx;y ð17Þ
4 : 2; : ;
1 5 τ6 τs
0 0 0 0 0
G12
8 9 8 9
> ϵ ϵ
< 1 > = < x >
> =
From the symmetry of the compliance matrix, then ϵ2 ϵy
¼T or in shortened form ϵ1;2 ¼ Tϵ x;y ð18Þ
>
: γ >
1 ; >
: γ >
1 ;
ν 12 ν 21 ν 13 ν 31 ν 23 ν 32 2 6
2 s
¼ ; ¼ ; ¼ ð13Þ
E1 E2 E 1 E3 E2 E3
Where the transformation matrix
Where Ei is the Young's modulus along axis i, Gij is the shear modulus 2 3
in direction j on the plane whose normal is in direction i, νij is Poisson's cθ 2 sθ 2 2cθ sθ
ratio for the i-j plane, σ1 , σ2 & σ3 are the normal stresses in directions 1, 2 T ¼ 4 sθ 2 cθ 2 −2cθ sθ 5; cθ ¼ cosθ; sθ ¼ sinθ ð19Þ
& 3 and τ4 , τ5 & τ6 are the shear stresses in 2–3, 1–3 and 1–2 planes −cθ sθ cθ sθ cθ 2 −sθ 2
respectively.
In the current applications, composite materials are used in the form σx;y ¼ Q x;y ϵ x;y ð20Þ
of thin laminates loaded in the plane of the laminate. The stress-strain
relations for a thin lamina (plane stress condition) can be expressed as Using Eqs. (15) to (20) thus,
followings: 2 3
Q 11 Q 12 0
2 3
1 ν 21 Q x;y ¼ T −1 4 Q 12 Q 22 0 5T ð21Þ
8 9 6 E − 0 78 9 0 0 2Q 66
E2
< ϵ1 = 6 1
6 ν 12 1
7 < σ1 =
7
ϵ2 ¼ 6
6 − 7
0 7 σ2 ð14Þ
: ; 6 E1 E2 7 : τ6 ; After dividing the third column of, Qx ,y, by 2, the laminate stiffness
γ6 4 1 5 matrices can be expressed as follows:
0 0
G12
i
The extensional stiffness matrix A ¼ ∑ Q kx;y ðhk −hk−1 Þ; ð22Þ
k¼1

1 i  
2 2
The coupling stiffness matrix B ¼ ∑ Q kx;y hk −hk−1 ð23Þ
2 k¼1

1 i  
3 3
The bending stiffness matrix D ¼ ∑ Q k h −hk−1 ð24Þ
3 k¼1 x;y k

Where, hk and hk-1 are the z-coordinates of the upper and lower sur-
faces of layer k measured from the laminate reference plane and i is the
number of layers (plies) in the basic laminate unit, see Fig. 5.
Then, the laminate compliance matrices can be obtained as follows:
n  o 
a ¼ A−1 − −A−1 B d BA−1 ; ð25Þ
Fig. 4. Fiber orientation in composite PV.
A.M. Kamal et al. / Desalination 397 (2016) 126–139 131

The shear force and moments per unit length Ns =Mx = My = Ms = 0


8 o9
< ϵx =
The reference plane strains matrix ϵ o x;y ¼ ϵoy ¼ aN x;y þ bMx;y
: o;
γs
ð33Þ

8 9
< κx =
The reference plane curvatures matrix κx;y ¼ κy ¼ cN x;y þ dM x;y
: ;
κs
Fig. 5. Layers coordinate notation.
ð34Þ

 
Then, the layer strains referred to global axes (x, y) can be given as
b ¼ −A−1 B d; ð26Þ
8 9 8 o9 8 9
  < ϵ x = < ϵx = < κx =
o
c ¼ −d BA−1 ; ð27Þ ϵy ¼ ϵy þ z κy ð35Þ
: ; : o; : ;
γs γs κs
 n o −1
d ¼ D− BA−1 B ð28Þ Where, z = coordinate of layer mid-plane measured from the
laminate reference plane, see Fig. 5.
Thus the layer strains referred to the principal material axes 1, 2 can
The mechanical loading matrices are
be written as
8 9
< Nx = 8 9 8 9
> ϵ ϵ
The forces matrix N x;y ¼ Ny ; ð29Þ < 1 > = >
< x > =
: ; ϵ2 ϵy
Ns ¼T ð36Þ
> 1
: γ ; > >
: γ >
1 ;
2 6 2 s
8 9
< Mx = And the layer stresses referred to the principal material axes 1, 2 are
The moments matrix M x;y ¼ My ð30Þ
: ; 8 9 2 38 9
Ms 0 < ϵ1 =
< σ1 = Q 11 Q 12
σ ¼ 4 Q 12 Q 22 0 5 ϵ2 ð37Þ
: 2; : ;
pDi τ6 0 0 Q 66 γ6
Where; the longitudinal force per unit length Nx ¼ σ x h ¼ in N=mm;
4
ð31Þ
The design of composite PV is based on the Tsai-Wu failure criterion
with the following form
pDi
the hoop force per unit length N y ¼ σ y h ¼ in N=mm: ð32Þ
2 f 1 σ 1 þ f 2 σ 2 þ f 11 σ 21 þ f 22 σ 22 þ f 66 τ26 þ 2f 12 σ 1 σ 2 ¼ 1 ð38Þ

Fig. 6. Finite element model for stainless steel PVs. (a) 2.5 in., (b) 4 in., (c) 8 in. and (d) 16 in.
132 A.M. Kamal et al. / Desalination 397 (2016) 126–139

Table 4 Table 5
FEM information. Element size in (PVs) FEM.

Geometry Surface body Solving time (sec)


PV size Element size
Stiffness behavior Flexible in mm Total number of elements Stainless steel Composite
Thickness mode Manual
2.5 2.5 12,960 11.357 52.167
Analysis type 3-D
4 4 20,240 16.037 118.436
Coordinate system Global Cartesian (x, y, z)
8 8 10,160 9.095 111.759
Mesh 16 16 5120 5.288 110.386

Physics preference Mechanical


Element type Linear quadrilateral (SHELL181)
Smoothing Medium Solving Eq. (40), the layer safety factor is
Mapped face meshing method Quadrilaterals
Body sizing type Element size qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Behavior Soft −β þ β2 þ 4α
Sfk ¼ ð42Þ
Solution 2α
Physics Structural
Analysis Static structural Using the first-ply failure (FPF) approach, the laminate safety factor
Solver Mechanical APDL equals to the minimum value of layer safety factor, Sfmin.
Environment temperature 22 °C The allowable PV thickness, ha in mm, can be determined as
followings:

Where, the Tsai-Wu coefficients can be calculated as followings: Sfall


ha ¼ ho ð43Þ
Sfmin
1 1 1 1
f1 ¼ − ; f2 ¼ −
F 1t F 1c F 2t F 2c qffiffiffiffiffiffiffiffiffiffiffiffiffiffi where; ho is the thickness of the basic laminate unit in mm ðho ¼ i  t Þ
1 1 1 1
f 11 ¼ ; f 22 ¼ ; f 66 ¼ 2 ; f 12 ¼ − f 11 f 22 ð44Þ
F 1t F 1c F 2t F 2c F6 2
ð39Þ t is the ply thickness = 0.2 mm,
Sfall = 2 (the same value for stainless steel PV).
where f1 & f2 are in MPa−1, f11 ,f22 , f66 & f12 are in MPa−2. The number of basic laminates, n can be calculated and approximat-
Then, this failure criterion is applied for each layer at failure stresses ed to the higher integer number as follows:
Sfk σ1, Sfk σ2and Sfk τ6 where, Sfk is the safety factor of layer k. Substitu-
tion of these failure stresses in Eq. (38) yields ha Sfall
n¼ ¼ ð45Þ
ho Sfmin
αS2fk þ βSfk −1 ¼ 0 ð40Þ
Then, the PV thickness is
where; α ¼ f 11 σ 21 þ f 22 σ 22 þ f 66 τ 26 þ 2f 12 σ 1 σ 2 ; β ¼ f 1 σ 1 þ f 2 σ 2
ð41Þ h ¼ n  ho ð46Þ

Fig. 7. Variation of safety factor versus element size for stainless steel PVs. (a) 2.5 in., (b) 4 in., (c) 8 in. and (d) 16 in.
A.M. Kamal et al. / Desalination 397 (2016) 126–139 133

And the PV safety factor is

h
Sf ¼ S ð47Þ
ha fall

The weight of composite PV, Wcomp in kg, is

π h i
2
W comp ¼ ρcomp ðDi þ 2hÞ −D2i L ð48Þ
4

where, ρcomp. is the density of composite material in kg/m3.


In case of hybrid composite material, the weight of PV can be deter-
Fig. 8. Distribution of the von-Mises stress for 8 in. PV.
mined as follows:

π ρg þ ρc h 2
i 3.2. Modeling of composite PV
W comp ¼ Di þ 2h −D2i L at ng ¼ nc ð49Þ
4 2
The finite element model for composite PVs for different lay-ups
where, given in Table 3 has been performed at the optimum lay-ups resulting
from the analytical solution. The same FEM information of stainless
ρg. is the density of E-glass/epoxy material, steel PVs is used.
ρc. is the density of carbon/epoxy material, The selected element size and total number of elements for both
ng. is the number of E-glass/epoxy layers, stainless steel and composite PVs are shown in Table 5. Also, the table
nc. is the number of carbon/epoxy layers, shows the solving time of both stainless steel and composite (for exam-
ple, E-glass/epoxy [±ϴ]ns lay-up) PVs.
and
4. Results and discussion
π 2ρg þ ρc h 2
i
W comp ¼ Di þ 2h −D2i L at ng ¼ 2nc ð50Þ 4.1. Stainless steel PV
4 3

The wall thickness and weight of stainless steel PVs of different sizes
for both analytical solution and FEM are shown in Table 6. It is clear that
3. Finite element modeling
the safety factor obtained from the FEM is slightly larger than that ob-
tained from the analytical solution for the same vessel wall thickness
The finite element modeling technique has been established for the
and size.
SWRO (PV) at the same sizes, materials and loading conditions that
As could be observed from Table 6, the FEM results are in good
used in the analytical solution to verify the results. ANSYS Workbench
agreement with the analytical solution within an acceptable range of
version 15 mechanical module is used in case of stainless steel PV and
the safety factor error i.e. b 0.4%. The distributions of the equivalent
ANSYS Composite PrepPost (ACP) module is used in case of composite
(von-Mises) stress and the safety factor for 8 in. PV are shown in
PV.
Figs. 8 and 9 respectively. From these figures, the maximum stress is
103.11 MPa and the minimum safety factor is 2.0076.
3.1. Modeling of stainless steel PV

The finite element model for stainless steel PVs of different sizes are 4.2. Composite PV
shown in Fig. 6. A DELL-Laptop-Core i5 has CPU speed of 2.5 GHz and
RAM of 4 GB with Windows 7-64 bit operating system is used. The The analysis of composite PVs is performed for all vessel sizes with
FEM information used for all vessel sizes is shown in Table 4. the different materials and lay-ups shown in Table 3. The variation of
vessel thickness with fiber orientation is plotted for each vessel and
3.1.1. Mesh convergence study the optimum winding angle at minimum thickness is determined.
Since the computed results of the FEM are affected by mesh size Then, the weight saving of each composite PV compared with the stain-
[18–21], a mesh convergence study has been performed to determine less steel one is calculated. FEM results are compared with that of the
the element size for various vessel sizes. The variation of PV safety factor analytical solution and the safety factor error is determined for each
with element size for all vessel sizes is shown in Fig. 7. As shown from vessel.
this figure, the safety factor decreases with decreasing of the element
size and close to the allowable value (Sfall = 2) for all vessel sizes. It is
noticed that the solving time increases with decreasing of the element
size, so that a suitable element size is selected for each vessel size.

Table 6
Wall thickness and weight of stainless steel PVs.

PV size h Weight Analytical solution FEM


inch mm kg Sf Sf ΔSf %

2.5 2.125 1.359 2 2.0073 0.365


4 3.401 8.697 2 2.0079 0.395
8 6.801 34.787 2 2.0076 0.380
16 13.602 139.150 2 2.0076 0.380
Fig. 9. Distribution of the safety factor for 8 in. PV.
134 A.M. Kamal et al. / Desalination 397 (2016) 126–139

Fig. 10. Variation of allowable thickness versus fiber orientation for (a) 2.5 in., (b) 4 in., (c) 8 in. and (d) 16 in. PVs for [±ϴ]ns lay-up.

4.2.1. Analytical solution materials depending on the amount of each material and the arrange-
The variation of the allowable PV thickness with fiber orientation ment of layers through the vessel wall thickness. The number of
(angle, ϴ) of different vessel sizes for [± ϴ]ns lay-up is shown in E-glass/epoxy layers is taken to be the same as carbon/epoxy layers.
Fig. 10. From this figure, it is clear that the allowable thickness depends The minimum allowable thickness is achieved at the optimum fiber ori-
on fiber orientation and material type. The carbon/epoxy vessel has al- entation of 56° for [+ϴC/−ϴG]ns and 53° for [±ϴC/±ϴG]ns lay-ups for
lowable thickness lower than the E-glass/epoxy vessel at the same all vessel sizes. The difference between vessel thicknesses for hybrid
winding angle for all vessel sizes. This is due to the high stiffness and lay-ups decreases with the variation of fiber orientation far from the op-
strength of the carbon fibers compared with the E-glass fibers. It is timum angle for all vessel sizes due to the little effect of carbon fibers
also observed that the minimum allowable thickness is achieved at stiffness and strength at the extreme values of winding angle (same
the optimum fiber orientation of 54° for E-glass/epoxy vessel and 55° thicknesses at ϴ = 0° and at ϴ = 90°) as shown in Fig. 10. For example,
for carbon/epoxy vessel for all vessel sizes. These results show good the minimum allowable thickness for [+ ϴC/− ϴG]ns lay-up is
agreement with the previous works [6,7,9,14]. 8.177 mm while for [± ϴC/± ϴG]ns lay-up is 4.592 mm for 8 in. PV.
In case of hybrid composite materials, the allowable thickness This is due to the existence of carbon fibers in helical direction for
values are between that of E-glass/epoxy and carbon/epoxy composite both positive and negative values of ϴ in [± ϴC/± ϴG]ns lay-up

Fig. 11. Variation of allowable thickness versus fiber orientation for (a) 2.5 in., (b) 4 in., (c) 8 in. and (d) 16 in. PVs for [0/±ϴ]ns lay-up.
A.M. Kamal et al. / Desalination 397 (2016) 126–139 135

Fig. 12. Variation of allowable thickness versus fiber orientation for (a) 2.5 in., (b) 4 in., (c) 8 in. and (d) 16 in. PVs for [90/±ϴ]ns lay-up.

(balanced laminate) which increase the strength and stiffness of the minimum allowable thickness for [0C/±ϴG]ns lay-up is 9.541 mm
whole vessel while for [+ ϴC/− ϴG]ns lay-up (unbalanced laminate), while for [0G/+ ϴG/− ϴC]ns lay-up is 7.814 mm for 8 in. PV. This is
the carbon fibers are existent in helical direction for positive values of due to the existence of carbon fibers in helical direction of negative
ϴ only. values of ϴ for [0G/+ ϴG/− ϴC]ns lay-up which increase the strength
The variation of the allowable PV thickness with fiber orientation of and stiffness of the whole vessel than that of the existence of carbon fi-
different vessel sizes for [0/± ϴ]ns lay-up is shown in Fig. 11. It is ob- bers in longitudinal direction only (angle, 0°) for [0C/±ϴG]ns lay-up.
served from this figure that the minimum allowable thickness is The variation of the allowable PV thickness with fiber orientation of
achieved at the optimum fiber orientation of 90° for E-glass/epoxy and different vessel sizes for [90/±ϴ]ns lay-up is shown in Fig. 12. It can be
carbon/epoxy vessels for all vessel sizes. For hybrid composite vessels, noticed from this figure that the minimum allowable thickness is
the number of E-glass/epoxy layers is taken to be twice the number of achieved at the optimum fiber orientation of 49° for E-glass/epoxy and
carbon/epoxy layers. The minimum allowable thickness is achieved at 46° for carbon/epoxy vessels for all vessel sizes. For hybrid composite
the optimum fiber orientation of 90° for [0C/± ϴG]ns and 79° for vessels, the number of E-glass/epoxy layers is taken to be twice the
[0G/+ ϴG/− ϴC]ns lay-ups for all vessel sizes. For example, the number of carbon/epoxy layers. The minimum allowable thickness is

Fig. 13. Variation of allowable thickness versus fiber orientation for (a) 2.5 in., (b) 4 in., (c) 8 in. and (d) 16 in. PVs for [0/±ϴ/90]ns lay-up.
136 A.M. Kamal et al. / Desalination 397 (2016) 126–139

Fig. 14. Variation of allowable thickness versus fiber orientation for (a) 2.5 in., (b) 4 in., (c) 8 in. and (d) 16 in. PVs for [90/±ϴ/90]ns lay-up.

achieved at the optimum fiber orientation of 26° for [90C/±ϴG]ns and carbon/epoxy layers. The minimum allowable thickness is achieved at
51° for [90G/+ ϴG/− ϴC]ns lay-ups for all vessel sizes. For example, the optimum fiber orientation of 90° for [0C/± ϴG/90C]ns and 61° for
the minimum allowable thickness for [90C/± ϴG]ns lay-up is [0G/±ϴC/90G]ns lay-ups for all vessel sizes. For example, the minimum
6.852 mm while for [90G/+ ϴG/− ϴC]ns lay-up is 8.794 mm for 8 in. allowable thickness for [0C/±ϴG/90C]ns lay-up is 4.958 mm while for
PV. This is due to the existence of carbon fibers in hoop direction [0G/±ϴC/90G]ns lay-up is 5.572 mm for 8 in. PV. This is due to the exis-
(angle, 90°) for [90C/± ϴG]ns lay-up which increase the strength and tence of carbon fibers in both hoop and longitudinal directions for
stiffness of the whole vessel than that of the existence of carbon fibers [0C/± ϴG/90C]ns lay-up while they existent in helical direction only
in helical direction of negative values of ϴ for [90G/+ϴG/− ϴC]ns lay- for [0G/±ϴC/90G]ns lay-up.
up. This leads to that the carbon fibers are more effective in hoop direc- The variation of the allowable PV thickness with fiber orientation of
tion which has higher stresses than other directions. These results show different vessel sizes for [90/±ϴ/90]ns lay-up is shown in Fig. 14. This
good agreement with the previous work [8]. figure demonstrates that the minimum allowable thickness is achieved
The variation of the allowable PV thickness with fiber orientation of at the optimum fiber orientation of 44° for E-glass/epoxy and 38° for
different vessel sizes for [0/±ϴ/90]ns lay-up is shown in Fig. 13. From carbon/epoxy vessels for all vessel sizes. In case of hybrid composite
this figure, it is obvious that the minimum allowable thickness is vessels, the number of E-glass/epoxy layers is taken to be the same as
achieved at the optimum fiber orientation of 90° for E-glass/epoxy and carbon/epoxy layers. The minimum allowable thickness is achieved at
71° for carbon/epoxy vessels for all vessel sizes. For hybrid composite the optimum fiber orientation of 0° for [90C/± ϴG/90C]ns and 50° for
vessels, the number of E-glass/epoxy layers is taken to be the same as [90G/± ϴC/90G]ns lay-ups for all vessel sizes. For example, the

Table 7
Wall thickness, weight and safety factor for 2.5 in. PVs.

Analytical solution FEM


Lay-up Material Optimum lay-up n ha h Weight Weight saving h ΔSf %
mm mm Sf kg % mm Sf

[±ϴ]ns E-glass/epoxy [±54]4s 4 2.923 3.2 2.1898 0.563 58.537 3.2 2.1895 −0.012
Carbon/epoxy [±55]2s 2 0.847 1.6 3.7790 0.207 84.776 1.6 3.7801 0.029
[+ϴC/−ϴG]ns [+56C/−56G]4s 4 2.555 3.2 2.5047 0.494 63.671 3.2 2.4498 −2.192
[±ϴC/±ϴG]ns [±53C/±53G]s 1 1.435 1.6 2.2301 0.241 82.271 1.6 2.2294 −0.031
[0/±ϴ]ns E-glass/epoxy [0/±90]4s 4 3.921 4.8 2.4483 0.865 36.314 4.8 2.4501 0.072
Carbon/epoxy [0/±90]s 1 0.942 1.2 2.5485 0.154 88.652 1.2 2.5502 0.068
[0C/±ϴG]ns [0C/±90G]3s 3 2.981 3.6 2.4149 0.585 56.948 3.6 2.4169 0.081
[0G/+ϴG/−ϴC]ns [0G/+79G/−79C]3s 3 2.442 3.6 2.9484 0.585 56.948 3.6 2.9102 −1.296
[90/± ϴ]ns E-glass/epoxy [90/±49]3s 3 3.315 3.6 2.1720 0.638 53.074 3.6 2.1722 0.008
Carbon/epoxy [90/±46]s 1 0.894 1.2 2.6837 0.154 88.652 1.2 2.6842 0.020
[90C/±ϴG]ns [90C/±26G]2s 2 2.141 2.4 2.2417 0.383 71.812 2.4 2.2427 0.045
[90G/+ϴG/ϴC]ns [90G/+51G/−51C]3s 3 2.748 3.6 2.6200 0.585 56.948 3.6 2.5744 −1.739
[0/±ϴ/90]ns E-glass/epoxy [0/±90/90]3s 3 3.659 4.8 2.6234 0.865 36.314 4.8 2.6253 0.072
Carbon/epoxy [0/±71/90]s 1 0.918 1.6 3.4845 0.207 84.776 1.6 3.4869 0.068
[0C/±ϴG/90C]ns [0C/±90G/90C]s 1 1.549 1.6 2.0655 0.241 82.271 1.6 2.0666 0.053
[0G/±ϴC/90G]ns [0G/±61C/90G]2s 2 1.797 3.2 3.5607 0.494 63.671 3.2 3.5639 0.089
[90/± ϴ/90]ns E-glass/epoxy [90/±44/90]3s 3 3.465 4.8 2.7704 0.865 36.314 4.8 2.7709 0.019
Carbon/epoxy [90/±38/90]s 1 0.912 1.6 3.5081 0.207 84.776 1.6 3.5093 0.034
[90C/±ϴG/90C]ns [90C/±0G/90C]2s 2 1.735 3.2 3.6892 0.494 63.672 3.2 3.6883 −0.024
[90G/±ϴC/90G]ns [90G/±50C/90G]s 1 1.324 1.6 2.4171 0.241 82.271 1.6 2.4135 −0.150
A.M. Kamal et al. / Desalination 397 (2016) 126–139 137

Table 8
Wall thickness, weight and safety factor for 4 in. PVs.

Analytical Solution FEM


Lay-up Material Optimum lay-up n ha h Weight Weight saving h ΔSf %
mm mm Sf kg % mm Sf

[±ϴ]ns E-glass/epoxy [±54]6s 6 4.676 4.8 2.0529 3.371 61.241 4.8 2.0527 −0.009
Carbon/epoxy [±55]2s 2 1.355 1.6 2.3620 0.820 90.572 1.6 2.3626 0.027
[+ϴC/−ϴG]ns [+56C/−56G]6s 6 4.088 4.8 2.3481 2.953 66.040 4.8 2.2991 −2.088
[±ϴC/±ϴG]ns [±53C/±53G]2s 2 2.296 3.2 2.7876 1.939 77.701 3.2 2.7867 −0.033
[0/±ϴ]ns E-glass/epoxy [0/±90]6s 6 6.274 7.2 2.2953 5.170 40.551 7.2 2.2967 0.061
Carbon/epoxy [0/±90]2s 2 1.507 2.4 3.1856 1.239 85.748 2.4 3.1877 0.066
[0C/±ϴG]ns [0C/±90G]4s 4 4.770 4.8 2.0124 3.093 64.441 4.8 2.0141 0.082
[0G/+ϴG/−ϴC]ns [0G/+79G/−79C]4s 4 3.907 4.8 2.4570 3.093 64.441 4.8 2.4280 −1.180
[90/± ϴ]ns E-glass/epoxy [90/±49]5s 5 5.304 6 2.2625 4.261 51.005 6 2.2626 0.003
Carbon/epoxy [90/±46]2s 2 1.431 2.4 3.3546 1.239 85.748 2.4 3.3554 0.025
[90C/±ϴG]ns [90C/±26G]3s 3 3.426 3.6 2.1016 2.293 73.631 3.6 2.1026 0.049
[90G/+ϴG/−ϴC]ns [90G/+51G/−51C]4s 4 4.397 4.8 2.1833 3.093 64.441 4.8 2.1488 −1.581
[0/±ϴ/90]ns E-glass/epoxy [0/±90/90]4s 4 5.855 6.4 2.1862 4.562 47.545 6.4 2.1875 0.060
Carbon/epoxy [0/±71/90]s 1 1.469 1.6 2.1778 0.820 90.572 1.6 2.1790 0.053
[0C/±ϴG/90C]ns [0C/±90G/90C]2s 2 2.479 3.2 2.5819 1.939 77.701 3.2 2.5838 0.075
[0G/±ϴC/90G]ns [0G/±61C/90G]2s 2 2.876 3.2 2.2254 1.939 77.701 3.2 2.2275 0.092
[90/± ϴ/90]ns E-glass/epoxy [90/±44/90]4s 4 5.544 6.4 2.3086 4.562 47.545 6.4 2.3091 0.020
Carbon/epoxy [90/±38/90]s 1 1.459 1.6 2.1926 0.820 90.572 1.6 2.1933 0.033
[90C/±ϴG/90C]ns [90C/±0G/90C]2s 2 2.776 3.2 2.3057 1.939 77.701 3.2 2.3051 −0.026
[90G/±ϴC/90G]ns [90G/±50C/90G]2s 2 2.118 3.2 3.0214 1.939 77.701 3.2 3.0167 −0.156

minimum allowable thickness for [90C/±ϴG/90C]ns lay-up is 5.551 mm From these tables, it is obvious that the lowest value of the minimum al-
while for [90G/±ϴC/90G]ns lay-up is 4.236 mm for 8 in. PV. This is due lowable thickness is achieved at the [±54]ns lay-up of the E-glass/epoxy
to the existence of carbon fibers in hoop direction only for [90C/±ϴG/ PVs and at the [±55]ns lay-up of the carbon/epoxy PVs while for hybrid
90C]ns lay-up while they existent in helical direction for [90G/± ϴC/ composite PVs, it is achieved at the [90G/±50C/90G]ns lay-up for all ves-
90G]ns which increase the strength and stiffness of the whole vessel. sel sizes.
From the previous analysis of the different types of hybrid composite The relative weight saving for each composite PV compared with the
lay-ups, it can be concluded that the existence of carbon fibers in helical stainless steel PV of the same size can be calculated as followings:
direction of both positive and negative values of ϴ is more effective in
increasing the strength and stiffness of the whole vessel than that of W st− W comp
Weight saving ¼ ð51Þ
the existence of them in hoop direction only but less effective than W st
both hoop and longitudinal directions. Also, the existence of carbon fi-
bers in helical direction of positive or negative values of ϴ is more effec- It is clear from Tables 7–10 that the carbon/epoxy PVs have weight
tive than the longitudinal direction only but less effective than hoop savings higher than that of the E-glass/epoxy PVs, good agreement
direction only. with the previous work [14], and the hybrid composite PVs have values
The minimum allowable thickness, thickness, safety factor, weight between them for all lay-ups and vessel sizes. For example, the opti-
and relative weight saving of 2.5 in., 4 in., 8 in. and 16 in. composite mum lay-up, [±55]4s, for carbon/epoxy 8 in. vessel has weight saving
PVs at the optimum lay-ups are shown in Tables 7–10 respectively. of 90.572% while [± 53C/±53G]3s lay-up has weight saving of

Table 9
Wall thickness, weight and safety factor for 8 in. PVs.

Analytical Solution FEM


Lay-up Material Optimum lay-up n ha h Weight Weight saving h ΔSf %
mm mm Sf kg % mm Sf

[±ϴ]ns E-glass/epoxy [±54]12s 12 9.353 9.6 2.0529 13.483 61.242 9.6 2.0527 −0.009
Carbon/epoxy [±55]4s 4 2.710 3.2 2.3620 3.280 90.572 3.2 2.3626 0.027
[+ϴC/−ϴG]ns [+56C/−56G]11s 11 8.177 8.8 2.1524 10.788 68.987 8.8 2.1100 −1.972
[±ϴC/±ϴG]ns [±53C/±53G]3s 3 4.592 4.8 2.0907 5.774 83.403 4.8 2.0900 −0.034
[0/±ϴ]ns E-glass/epoxy [0/±90]11s 11 12.547 13.2 2.1040 18.853 45.806 13.2 2.1056 0.075
Carbon/epoxy [0/±90]3s 3 3.014 3.6 2.3892 3.697 89.373 3.6 2.3905 0.055
[0C/±ϴG]ns [0C/±90G]8s 8 9.541 9.6 2.0124 12.370 64.441 9.6 2.0141 0.082
[0G/+ϴG/−ϴC]ns [0G/+79G/−79C]7s 7 7.814 8.4 2.1499 10.763 69.061 8.4 2.1254 −1.138
[90/± ϴ]ns E-glass/epoxy [90/±49]9s 9 10.608 10.8 2.0363 15.254 56.151 10.8 2.0363 0.001
Carbon/epoxy [90/±46]3s 3 2.862 3.6 2.5159 3.697 89.373 3.6 2.5165 0.023
[90C/±ϴG]ns [90C/±26G]6s 6 6.852 7.2 2.1016 9.173 73.631 7.2 2.1026 0.049
[90G/+ϴG/−ϴC]ns [90G/+51G/−51C]8s 8 8.794 9.6 2.1833 12.370 64.441 9.6 2.1452 −1.745
[0/±ϴ/90]ns E-glass/epoxy [0/±90/90]8s 8 11.710 12.8 2.1862 18.248 47.545 12.8 2.1875 0.060
Carbon/epoxy [0/±71/90]2s 2 2.939 3.2 2.1778 3.280 90.572 3.2 2.1791 0.058
[0C/±ϴG/90C]ns [0C/±90G/90C]4s 4 4.958 6.4 2.5819 7.757 77.701 6.4 2.5838 0.075
[0G/±ϴC/90G]ns [0G/±61C/90G]4s 4 5.752 6.4 2.2254 7.757 77.701 6.4 2.2275 0.092
[90/± ϴ/90]ns E-glass/epoxy [90/±44/90]7s 7 11.089 11.2 2.0201 15.848 54.442 11 2.0205 0.022
Carbon/epoxy [90/±38/90]2s 2 2.919 3.2 2.1926 3.280 90.572 3.2 2.1933 0.033
[90C/±ϴG/90C]ns [90C/±0G/90C]4s 4 5.551 6.4 2.3057 7.757 77.701 6.4 2.3052 −0.022
[90G/±ϴC/90G]ns [90G/±50C/90G]3s 3 4.236 4.8 2.2660 5.774 83.403 4.8 2.2625 −0.157
138 A.M. Kamal et al. / Desalination 397 (2016) 126–139

Table 10
Wall thickness, weight and safety factor for 16 in. PVs.

Analytical Solution FEM


Lay-up Material Optimum lay-up n ha h Weight Weight saving h ΔSf %
mm mm Sf kg % mm Sf

[±ϴ]ns E-glass/epoxy [±54]24s 24 18.705 19.2 2.0529 53.932 61.242 19.2 2.0525 −0.019
Carbon/epoxy [±55]7s 7 5.419 5.6 2.0667 11.457 91.766 5.6 2.0672 0.023
[+ϴC/−ϴG]ns [+56C/−56G]21s 21 16.353 16.8 2.0546 41.115 70.453 16.8 2.0156 −1.899
[±ϴC/±ϴG]ns [±53C/±53G]6s 6 9.183 9.6 2.0907 23.094 83.403 9.6 2.0899 −0.039
[0/±ϴ]ns E-glass/epoxy [0/±90]21s 21 25.095 25.2 2.0084 71.784 48.413 25.2 2.0099 0.075
Carbon/epoxy [0/±90]6s 6 6.027 7.2 2.3892 14.787 89.373 7.2 2.3908 0.067
[0C/±ϴG]ns [0C/±90G]16s 16 19.081 19.2 2.0124 49.480 64.441 19.2 2.0143 0.092
[0G/+ϴG/−ϴC]ns [0G/+79G/−79C]14s 14 15.629 16.8 2.1499 43.051 69.061 16.8 2.1242 −1.194
[90/± ϴ]ns E-glass/epoxy [90/±49]18s 18 21.215 21.6 2.0363 61.016 56.151 21.6 2.0362 −0.004
Carbon/epoxy [90/±46]5s 5 5.724 6 2.0966 12.287 91.170 6 2.0971 0.024
[90C/±ϴG]ns [90C/±26G]12s 12 13.704 14.4 2.1016 36.692 73.631 14.4 2.1026 0.049
[90G/+ϴG/−ϴC]ns [90G/+51G/−51C]15s 15 17.588 18 2.0469 46.257 66.757 18 2.0121 −1.698
[0/±ϴ/90]ns E-glass/epoxy [0/±90/90]15s 15 23.420 24 2.0495 68.175 51.006 24 2.0510 0.071
Carbon/epoxy [0/±71/90]4s 4 5.877 6.4 2.1778 13.119 90.572 6.4 2.1793 0.067
[0C/±ϴG/90C]ns [0C/±90G/90C]7s 7 9.915 11.2 2.2591 27.047 80.563 11.2 2.2608 0.074
[0G/±ϴC/90G]ns [0G/±61C/90G]8s 8 11.503 12.8 2.2254 31.029 77.701 12.8 2.2272 0.079
[90/± ϴ/90]ns E-glass/epoxy [90/±44/90]14s 14 22.178 22.4 2.0201 63.394 54.442 22.4 2.0204 0.017
Carbon/epoxy [90/±38/90]4s 4 5.838 6.4 2.1926 13.119 90.572 6.4 2.1933 0.033
[90C/±ϴG/90C]ns [90C/±0G/90C]7s 7 11.103 11.2 2.0175 27.047 80.563 11.2 2.0170 −0.025
[90G/±ϴC/90G]ns [90G/±50C/90G]6s 6 8.473 9.6 2.2660 23.094 83.403 9.6 2.2625 −0.157

83.403%. Then, the hybrid composite material which saves about 50% The distribution of the safety factor for [±54]12s lay-up 8 in. E-glass/
carbon fibers has weight saving close to the carbon/epoxy composite epoxy PV is shown in Fig. 15.
material with deference of 7.169%. Also, the optimum lay-up, [0/±
90]3s, for carbon/epoxy 8 in. vessel has weight saving of 89.373% while 5. Conclusion
[0G/+79G/−79C]7s lay-up has weight saving of 69.061%. Then, the hy-
brid composite material which saves about 66.6% carbon fibers has The SWRO pressure vessels (PVs) have been modeled using analyt-
weight saving deference of 20.312% from the carbon/epoxy composite ical solution and finite element modeling (FEM) for stainless steel and
material. Thus, the low cost of PVs can be achieved by using of the hy- fiber reinforced composite materials to optimize the PV design parame-
brid composite material for the same lay-up and vessel size. ters. The used PV sizes are 2.5, 4, 8 and 16 in. according to the standard
sizes in the international markets. For stainless steel PVs, the same wall
4.2.2. FEM solution thickness is obtained from both analytical solution and FEM for each
The PV thickness and safety factor obtained from the FEM for 2.5 in., vessel size with acceptable safety factor error.
4 in., 8 in. and 16 in. composite PVs at the optimum lay-ups are shown in For composite PVs, the analytical solution results show that the al-
Tables 7–10 respectively. The same vessel wall thickness is obtained as lowable PV thickness depends on fiber orientation and material type
in the analytical solution with some safety factor error for each vessel. for a certain lay-up. The carbon/epoxy vessels have allowable thickness
As could be observed from these tables, the safety factors obtained lower than the E-glass/epoxy vessels at the same winding angle for all
from the finite element modeling are close to that obtained from the an- vessel sizes. This is due to the high stiffness and strength of the carbon
alytical solution within an acceptable range i.e. b0.2% for most lay-ups fibers compared with the E-glass fibers. In case of hybrid composite ma-
except that for hybrid lay-ups which have carbon fibers in one helical terials, the allowable thickness values are between that of E-glass/epoxy
direction only, the safety factor error increases to about 2.2%. and carbon/epoxy materials depending on the amount of each material

Fig. 15. Distribution of the safety factor for [±54]12s lay-up 8 in. E-glass/epoxy PV.
A.M. Kamal et al. / Desalination 397 (2016) 126–139 139

and the arrangement of layers through the vessel thickness. By compar- [7] L. Parnas, N. Katirci, Design of fiber-reinforced composite pressure vessels under
various loading conditions, Compos. Struct. 58 (2002) 83–95.
ing the different types of lay-ups, the optimum lay-ups are [±54]ns lay- [8] Y.J. Xie, H.G. Yan, Z.M. Liu, Buckling optimization of hybrid-fiber multilayer-sand-
up for the E-glass/epoxy PVs, [±55]ns lay-up for the carbon/epoxy PVs wich cylindrical shells under external lateral pressure, Compos. Sci. Technol. 56
and [90G/±50C/90G]ns lay-up for the hybrid composite PVs for all ves- (1996) 1349–1353.
[9] H.H. Mian, G. Wang, U.A. Dar, W. Zhang, Optimization of composite material system
sel sizes. Thus, for hybrid composite PVs, it is recommended to use car- and lay-up to achieve minimum weight pressure vessel, Appl. Compos. Mater. 20
bon fibers in helical direction of both positive and negative values of (2013) 873–889.
winding angle to increase the strength and stiffness of the whole vessel. [10] X. Wang, G. Liang, W. Zhang, N. Du, J. Yang, Study on properties of pressure vessels
based on CF/AF hybrid fiber reinforced epoxy composites, Sci. Eng. Compos. Mater.
The composite PVs show better weight savings compared with the 13 (2006) 131–138.
stainless steel PVs. The carbon/epoxy PVs have weight savings higher [11] B. Assam, M. Muhammad, M. Mokhtar, F. Kolkailah, A theoretical and design analy-
than that of the E-glass/epoxy PVs and the hybrid composite PVs have sis of the filament-wound composite pressure vessels, Sci. Eng. Compos. Mater. 4
(1995) 73–88.
values between them for all lay-ups and vessel sizes. Thus, the low
[12] S.H. Taghavian, J.E. Jam, M. Zabihpoor, M. Yousefzadeh, Development of a new
cost PVs can be achieved by using of the hybrid composite material for method for design of stiffened composite pressure vessel using lattice structures,
the same lay-up and vessel size. Finally, the FEM results show good Sci. Eng. Compos. Mater. 22 (2014) 459–468.
agreement with the analytical solution results. [13] M.H. Sabour, M.F. Foghani, Design of semi-composite pressure vessel using fuzzy
and FEM, Appl. Compos. Mater. 17 (2010) 175–182.
[14] I.M. Daniel, O. Ishai, Engineering Mechanics of Composite Materials, second ed. Ox-
References ford University Press, New York; Oxford, 2006.
[15] J. Olsson, Stainless steels for desalination plants, Desalination 183 (2005) 217–225.
[1] Y.M. Kim, S.J. Kim, Y.S. Kim, S. Lee, I.S. Kim, J.H. Kim, Overview of systems engineer- [16] ANSYS Workbench version 15, Engineering Data Material Library, 2013.
ing approaches for a large-scale seawater desalination plant with a reverse osmosis [17] R.M. Jones, Mechanics of Composite Materials, McGraw-Hill, New York, 1975.
network, Desalination 238 (2009) 312–332. [18] M. Rezakazemi, M. Shahverdi, S. Shirazian, T. Mohammadi, A. Pak, CFD simulation of
[2] M.C. Porter, Handbook of Industrial Membrane Technology, Noyes, Park Ridge, N.J., water removal from water/ethylene glycol mixtures by pervaporation, Chem. Eng. J.
1990 168 (2011) 60–67.
[3] A.A. Krikanov, Composite pressure vessels with higher stiffness, Compos. Struct. 48 [19] M.F. Gruber, U. Aslak, C. Hélix-Nielsen, Open-source CFD model for optimization of
(2000) 119–127. forward osmosis and reverse osmosis membrane modules, Sep. Purif. Technol. 158
[4] P. Xu, J.Y. Zheng, P.F. Liu, Finite element analysis of burst pressure of composite hy- (2016) 183–192.
drogen storage vessels, Mater. Des. 30 (2009) 2295–2301. [20] G. Venkatesan, N. Kulasekharan, S. Iniyan, Design and selection of curved vane
[5] D.-S. Son, S.-H. Chang, Evaluation of modeling techniques for a type III hydrogen demisters using Taguchi based CFD analysis, Desalination 354 (2014) 39–52.
pressure vessel (70 MPa) made of an aluminum liner and a thick carbon/epoxy [21] E. Xu, et al., Investigations on the applicability of hydrostatic bearing technology in a
composite for fuel cell vehicles, Int. J. Hydrog. Energy 37 (2012) 2353–2369. rotary energy recovery device through CFD simulation and validating experiment,
[6] A. Onder, O. Sayman, T. Dogan, N. Tarakcioglu, Burst failure load of composite pres- Desalination 383 (2016) 60–67.
sure vessels, Compos. Struct. 89 (2009) 159–166.

Вам также может понравиться