Chen - Ultimate Longitu.2008.JSR

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Journal of Ship Research, Vol. 52, No. 3, September 2008, pp.

184–193

Ultimate Longitudinal Strength of Ship Hulls of


Composite Materials

Nian-Zhong Chen and C. Guedes Soares


Centre for Marine Technology and Engineering, Technical University of Lisbon, Instituto Superior Técnico, Lisboa, Portugal

A progressive collapse analysis method is proposed to predict the ultimate longitu-


dinal strength of ship hulls of composite materials. The load–average strain curve
derived from a progressive failure nonlinear finite element analysis is adopted for
representing the behavior of each stiffened composite panel forming a hull cross
section. The bending moment of the ship hull under a prescribed curvature is
achieved by integrating the reaction force of each stiffened panel over a hull cross
section based on the load–average strain curves. The ultimate longitudinal strength
of a ship hull is obtained from the moment–curvature relationship of the ship hull,
which is established by imposing progressively increasing curvatures of a hull cross
section. An all-composite ship is analyzed as an application.

Keywords: composites; longitudinal strength; buckling

1. Introduction plating, and a progressive collapse analysis is performed based on


the assumption that the cross section remains plane after defor-
THE IMPROVED DESIGN, fabrication, and mechanical performance mation and each panel behaves according to its average stress–
of low-cost composites have led to the increase of sizes of com- average strain relationships.
posite ships, and the use of composite materials has become a After Smith, various research papers were published to develop
common practice in marine structures. Presently, some all- the accurate average stress–average strain relationship of each
composite naval ships have been constructed up to 80 to 90 m stiffened panel forming a hull cross section. For instance, Gordo et
long. Mouritz et al. (2001) predicted that hulls for mid-sized war- al. (1996) modeled this relationship with simple analytical formu-
ships, such as frigates that are typically 120 to 160 m long, may be las, which was shown to work well (Gordo & Guedes Soares
constructed of composite materials by 2020. With the increase of 1996).
the sizes of composite ships, the development of methods for The finite element method (FEM) is also a powerful tool to
prediction of the ultimate longitudinal strength of composite ships perform the progressive collapse analysis for ship hulls. ABS
is necessary. group (Chen et al. 1983) and DNV group (Valsgaard et al. 1991)
The first attempt to predict the ultimate strength of steel ships have adopted the FEM to carry out this collapse analysis. The
was made by Caldwell (1965). He introduced the fully plastic idealized structural unit method (ISUM) originally proposed by
bending moment of a cross section considering the influence of Ueda and Rashed (1984) is another effective alternative method to
yielding of all structural members. However, the strength reduc-
perform the progressive collapse analysis for ship hulls. However,
tion in individual members after they have attained their ultimate
as indicated in Yao et al. (2000) and Yao (2003), the hull girder is
strength locally as well as the time lag in collapse of individual
in general too large to perform the progressive collapse analysis
members was not considered. This problem was solved by Smith
using the common FEM, and ISUM needs to develop more ap-
(1977), who proposed an approach in which the cross section is
propriate elements. The Smith method is generally the most ef-
divided into small elements composed of a stiffener and attached
fective method among the aforementioned ones for prediction of
the ultimate longitudinal strength of steel ships.
Manuscript received at SNAME headquarters September 7, 2006; revised Although considerable research has been conducted on predic-
manuscript received August 10, 2007. tion of the ultimate strength of steel ships, little concern has been

184 SEPTEMBER 2008 0022-4502/08/5203-0184$00.45/0 JOURNAL OF SHIP RESEARCH


shown regarding ship hulls of composite materials. Chen et al.
(2003) proposed an analytical method to predict the ultimate lon-
t

t KL = t
tV t
T
BLt Dtt BLdtV

K =兰
gitudinal strength of a composite ship where the behavior of stiff-
ened composite panels was represented by a simplified analytical
t
t NL
t
tV t BTNLt␶ tt BNLdtV (2)

F=兰
formula. The work was further extended by Chen and Guedes
L ␶d V
t t
Soares (2007) to develop a simplified method for estimation of the t tV t BTt ˆ t
ultimate longitudinal strength of composite ships based on a so-
phisticated strength model of stiffened composite panels. where tt B L and tt B NL are the linear and nonlinear strain-
With use of the basic idea of the Smith method (Smith 1977), displacement transformation matrices at time t and referred to
this paper extends the work of Chen and Guedes Soares (2007) to configuration t; tD is the incremental stress-strain material prop-
develop a progressive collapse analysis method for prediction of erty matrix; t␶ and t␶ˆ are the matrix and vector of Cauchy stresses.
the ultimate longitudinal strength of ship hulls in composite ma- The degenerated shell element originally proposed by Ahmad et
terials. The behavior of stiffened composite panels forming a hull al. (1970) for linear analysis of thin shells is used here for non-
cross section are represented by so-called “load–average strain linear analysis. This element is degenerated from the three-
curves,” which are estimated by a progressive failure nonlinear dimensional solid element by imposing two constraints: (1) the
finite element analysis, and the ultimate longitudinal strength of a normals to the mid-surface remain straight after deformation, but
ship hull is achieved from the moment–curvature relationship of not necessarily normal to the mid-surface; and (2) the transverse
the ship hull, which is established by imposing progressively in- normal components of strain and, hence, stresses are ignored
creasing curvatures of a hull cross section. where the transverse direction is defined as the through-thickness
direction, namely the 3 direction in the principal material coordi-
nate. The resulting nonlinear formulations admit large displace-
2. Load–average strain curves ments and rotations of the shell element and small strains. The
detailed description of this element can be found in Bathe (1996).
Stiffened composite panels used in ship hulls are usually lam-
inated structures, and thus the material properties are often non- 2.2. Explicit through-thickness integration scheme
uniform over their cross sections. This leads to the nonuniform
stress distribution over their cross sections when they are under In the conventional degenerated shell element formulations for
axial compression or tension. In order to represent conveniently laminated structures, the linear and nonlinear strain incremental
the behavior of each stiffened panel forming the mid-ship section, stiffness matrices and the vector of nodal point forces equivalent
a load–average strain curve, which is defined as the relationship to the element stresses are expressed as (Chang & Swawami-
between the reaction force and the average strain of the cross phakdi 1981)
section of a stiffened panel subjected to end-shortening displace-
ments, is developed herein.
The load–average strain curve is established by a progressive
t
t KL 兺 兰 共 B 兲 共 D兲 共 B 兲 共d V兲
=
N
m=1
t
tV t
T
L m t
t
m t L m
t
m

failure nonlinear finite element analysis, in which the updated K = 兺 兰 共 B 兲 共 ␶兲 共 B 兲 共d V兲


N
t
t NL m=1
t
tV t
T
NL m
t t
m t NL m
t
m (3)
Lagrangian formulation with a degenerated shell element (Ahmad
F = 兺 兰 共 B 兲 共 ␶ˆ 兲 共d V兲
et al. 1970) is used for structural analysis and an explicit through- t N t T t t
thickness integration scheme developed by Yunus et al. (1989) for t m=1 tV t L m m m

linear problems is extended to nonlinear analysis for improving


the computational efficiency. The Tsai-Wu criterion (Tsai & Wu where (ttBL)m, (ttBNL)m, (t␶)m, and (t␶ˆ )m are the linear and nonlinear
1971) is adopted to predict the failure mechanisms, and the stiff- strain-displacement transformation matrices, and the matrix and
ness reduction is carried out at the Gauss points of the finite the vector of Cauchy stresses of the mth ply, respectively. N is the
element mesh depending on the mode of failure. total number of plies. (tD)m is the incremental stress-strain mate-
rial property matrix.
Obviously, to get the linear and nonlinear strain incremental
2.1. Finite element formulation stiffness matrices and the vector of nodal point forces, three-
dimensional numerical integration in each ply has to be carried
Based on the principle of virtual work and the updated Lagran-
out, and thus the finite element analysis will become computa-
gian formulation, the final incremental equilibrium equations for a
tionally inefficient with the increase of the number of plies. In
single element of a continuous medium at time t + ⌬t are given by
order to overcome this problem, an explicit through-thickness in-
Bathe (1996):
tegration scheme developed by Yunus et al. (1989) for linear
problems is extended in the present method to nonlinear analysis,
共ttKL + ttKNL兲⌬U = t+⌬tR − ttF (1) in which the Jacobian inverse is assumed to vary linearly across
the thickness. Namely, the inverse of the Jacobian matrix is ap-
where ⌬U is the vector of increments in the nodal point displace-
proximately given by
ments; t+⌬tR is the vector of externally applied nodal point loads
at time t + ⌬t; ttKL and ttKNL are linear and nonlinear strain incre- t −1
= tJ−1 t −1
J a + ␨ Jd (4)
mental stiffness matrices at time t and referred to configuration t;
where ␨ is the natural coordinate in the thickness direction and
t
tF is the vector of nodal point forces equivalent to the element
stresses at time t.
t −1
t t t
tKL, tKNL, and tF can be written in the standard matrix form as Ja = 共tJ␨−1=1.0 + tJ␨−1=−1.0兲 Ⲑ 2

SEPTEMBER 2008 JOURNAL OF SHIP RESEARCH 185


t −1
Jd = 共tJ␨−1=1.0 − tJ␨−1=−1.0兲 Ⲑ 2 (5) based on Tsai-Wu criteria. If failure occurs, then the following
expressions are adopted to determine the failure mode
Also, the determinant of the Jacobian matrix is explicitly writ-
ten in the form H1 = F1␴1 + F11␴21, H2 = F2␴2 + F22␴22
H4 = F44␴12 H5 = F55␴223 H6 = F66␴213 (11)
ⱍJⱍ = 冋⌬ 冉 冊册
2
⌬d 2
t
1+␨ (6) The largest Hi term is selected as the dominant failure mode and
a
⌬a
the corresponding engineering properties of material at the Gauss
where points are reduced to zero. H1 corresponds to E1, H2 to E2, H4 to
G12, H5 to G23, H6 to G13, in which Ei is elastic modulus in
⌬a = 共 公ⱍJⱍ t
␨=1.0 + 公ⱍJⱍ t
␨=−1.0 兲Ⲑ 2 principal material direction i, and Gij is shear modulus in principal
material planes ij.
⌬d = 共公ⱍ Jⱍ t
␨=1.0 − 公ⱍ Jⱍ t
␨=−1.0 兲 Ⲑ 2 (7)
2.5. Procedure of progressive failure nonlinear finite
The linear and nonlinear strain-displacement transformation element analysis
matrices ttBL and ttBNL are thus given by
An outline of the steps used in the progressive failure nonlinear
t
t BL = tt BL1 + ␨ tt BL2 + ␨2 tt BL3 finite element analysis for establishing the load–average strain
curve of a stiffened composite panel is described. The load–
t
t BNL = tt BNL1 + ␨ tt BNL2 + ␨2 tt BNL3 (8) average strain curve consists of two parts. One part represents the
where t t t t t t behavior of the stiffened panel under compression, and the other
tBL1, tBL2, tBL3, tBNL1, tBNL2, and are the components
tBNL3
independent of ␨. Substitution of equation (8) into equation (3) and part represents the tension behavior of the stiffened panel. The two
taking the ply summation inside the surface integrations, the re- parts are both established by the following steps:
sultant ttKL, ttKNL, and ttF are then independent of ␨. Therefore, an 1. Define the initial state, including geometries, material proper-
explicit through-thickness integration can be performed for ttKL, ties, boundaries, and so forth.
t t
tKNL, and tF and the computational efficiency is improved. 2. At each end-shortening displacement, the nonlinear finite ele-
ment analysis is performed until convergence. Stresses in the
2.3. Failure identification global coordinates at the middle of each layer at each Gauss
point are calculated. Then those stresses are transformed to
In order to identify the material failure of the structure, the principal material coordinates.
Tsai-Wu criterion is adopted in the paper. Material failure is con- 3. Compute the failure index, F. If failure occurs (i.e., F ⱖ 1), (a)
sidered to have occurred if the following single condition is sat- identify the maximum Hi, (b) remove the corresponding engi-
isfied at any Gauss point of the structure: neering properties of the material at that Gauss point and un-
load the corresponding stresses instantaneously, and (c) recom-
F ≡ F1␴1 + F2␴2 + F11␴21 + F22␴22 + 2F12␴1␴2 + F44␴212 + F55␴223
pute the stiffness matrix of the structure and reestablish the
+ F66␴213 ⱖ 1 (9) equilibrium state at the same end-shortening displacement.
where 4. If no failure occurs, proceed to the next step of end-shortening
displacement until the collapse of the structure, that is, the
1 1 1 1 1 1 solution is diverged or the stiffness matrix of the structure
F1 = − , F2 = − , F11 = , F22 = , becomes singular.
XT XC YT YC XTXC YTYC
1 1 1 1 1 2.6. Numerical accuracy
F44 = 2, F55 = 2, F66 = 2, F12 = − (10)
R S T 2 公XTXCYTYC
The numerical accuracy of the progressive failure nonlinear
where ␴i denote the stress components referred to the principal finite element analysis is evaluated with reference to the experi-
material coordinates; XT and YT are the lamina normal strengths in mental data from Starnes and Rouse (1981) and Kong et al. (1998)
tension along material coordinates 1 and 2; XC and YC are the and also the corresponding numerical results from other authors
lamina normal strengths in compression along material coordi- (Engelstad et al. 1992, Sleight et al. 1997, Kong et al. 1998).
nates 1 and 2; R, S, and T are the lamina shear strengths in the A laminated composite plate denoted C4 from Starnes and
material planes 12, 23, 13, respectively. Rouse (1981) and a stiffened composite panel denoted S1 from
Kong et al. (1998) are utilized for this evaluation. C4 is a 24-ply
2.4. Property degradation model orthotropic laminate with a stacking sequence of [±45/02/±45/02/
±45/0/90]s. Its dimension is 50.8 cm long × 17.8 cm wide, and the
A limited discount model (Engelstad et al. 1992, Shahid & lamina thickness is 0.14 mm. C4 is made of graphite/epoxy ma-
Chang 1995, Sleight et al. 1997), in which the stiffness reduction terial the properties of which are shown in Table 1. It is loaded in
is carried out at the Gauss points of the finite element mesh de- axial compression, and boundaries are clamped for loaded edges
pending on the mode of failure, is used to simulate the progressive (along width) and simply supported along the unloaded sides
failure process of stiffened composite panels under axial compres- (along length).
sion or tension. S1 is a panel with blade-section stiffeners. Its geometries and
At each load step, the stresses at Gauss points of finite element dimensions are shown in Fig. 1. It is fabricated using graphite/
mesh are used to investigate the occurrence of material failure epoxy material the properties of which are shown in Table 1. S1

186 SEPTEMBER 2008 JOURNAL OF SHIP RESEARCH


Table 1 Material properties of C4 and S1 pressive strengths of C4 and S1 of the present model generally
approach those predicted by Engelstad et al. (1992), Sleight et al.
Property Symbol C4 S1 (1997), and Kong et al. (1998). Figure 3 shows that the predictions
of the present model correlate very well with the experimental data
Elastic modulus in 1 principal
material direction E1 131.0 GPa 130.0 GPa of C4 and S1. Also, Fig. 3 shows that the agreement between the
Elastic modulus in 2 principal experimental data of S1 and the estimated results of the present
material directions E2 13.0 GPa 10.0 GPa model is much better than that reported in Kong et al. (1998).
Shear modulus in 1–2 and 1–3 It might be suggested from Table 2 and Fig. 3 that the progres-
principal material planes G12, G13 6.4 GPa 4.85 GPa sive failure nonlinear finite element analysis can provide a rea-
Shear modulus in 2–3 principal sonable prediction of load–average strain curves of stiffened com-
material planes G23 1.7 GPa 3.62 GPa posite panels.
Poisson’s ratio v12 0.38 0.31
Tensile strength in 1 principal
material direction XT 1.4 GPa 1.933 GPa 3. Progressive collapse analysis
Compressive strength in 1
principal material direction XC 1.138 GPa 1.051 GPa When the load–average strain curve of each stiffened panel
Tensile strength in 2 principal forming the cross section of the mid-ship is available, the ultimate
material directions YT 80.9 MPa 51 MPa longitudinal strength of the ship hull is achieved by a progressive
Compressive strength in 2 collapse analysis in which the moment–curvature relationship of
principal material directions YC 189 MPa 141 MPa
the ship hull is established by imposing progressively increasing
Shear strength in 1–2 principal
curvatures and determining the associated bending moments.
material planes R 69 MPa 61 MPa
Shear strength in 2–3 principal
material planes S 62 MPa 61 MPa 3.1. Hypothesis
Shear strength in 1–3 principal
material planes T 62 MPa 61 MPa To perform the progressive collapse analysis for the ship hull,
with use of the basic idea of the Smith method (Smith 1977),
several basic hypotheses are made as
consists of two stiffeners with the lay-up sequences of [0/90/45/ • The cross section of the mid-ship is divided into a series of
0/−45]s, and the stiffeners are formed by continuous lay-up of the stiffened panels, and each of them is considered to act and
skin [0/90/±45]s. It is subjected to axial compressive loads, and behave independently.
boundaries are clamped for loaded edges (along width) and free • The cross section of the mid-ship remains plane, the normal
along the unloaded sides (along length). strain varies linearly across the perpendicular direction of the
Finite element models of C4 and S1 are both developed by cross section due to hull girder flexure, and the influence of
nine-node quadratic degenerated laminated shell elements. Due to transverse restraint on longitudinal stress (Poisson’s ratio ef-
structural symmetry, only one-half of them are modeled, as shown fect) is negligible.
in Fig. 2. • Longitudinal collapse occurs only between two adjacent
After analysis, the comparison between the experimental data transverse frames.
and the estimated results of the present model and the references
is shown in Table 2. The graphs of the estimated end-shortening 3.2. Instantaneous neutral axis
versus the applied load comparing with experimental data are
shown in Fig. 3. The first step of the progressive collapse analysis is to estimate
Table 2 shows that the relative errors between the experimental the position of the neutral axis of the mid-ship section under a
results of C4 and S1 and the estimated results of the present model prescribed curvature. Here a trial-and-error process is used to es-
are both less than 5%, and also the estimated postbuckling com- timate the position of the neutral axis. According to the force

Fig. 1 Geometries and dimensions of S1

SEPTEMBER 2008 JOURNAL OF SHIP RESEARCH 187


Table 2 Comparison between estimated results and
experimental data (knots)

Model Failure Criteria C4 S1

Present model Tsai-Wu 101.3 27.7


Maximum stress 111.4 —
Engelstad et al. (1992) Tsai-Wu 104.3 —
Christensen 99.8 —
Sleight et al. (1997) Hashin 104.6 —
Kong et al. (1998) Maximum stress — 27.4
Experimental data 98.0 26.8

strain varies linearly across the perpendicular direction of the mid-


ship section, the average strain ␧i of stiffened panel i is approxi-
mately given by
␧i = zi␾
zi = Zi − ZNA (13)
where ␾ is the prescribed curvature of the mid-ship section and Zi
is the vertical coordinate of the neutral axis of stiffened panel i,
which is given by

Z =
兺 N
j=1EjAjYj
(14)

i N
j=1EjAj

where Aj, Ej, and Yj represent the cross section area, Young’s
modulus, and the vertical coordinate of the jth element in the
stiffened panel i, respectively. N is the total number of elements in
the stiffened panel i.
Once the average strain ␧i is given by equation (13), the corre-
sponding reaction force Fi over the cross section of the stiffened
panel i can be obtained from its load–average strain curve. Then,
F is achieved by integrating each Fi over the mid-ship section.
In the trial-and-error process, F is regarded as the value of
evaluation function for each individual ZNA. When F is not close
enough to 0, ZNA is necessary to shift according to the sign of F.
The final solution of ZNA is reached when the convergence criteria
|F| < ⌬ (say 10−3) is satisfied.

3.3. Ultimate longitudinal strength


When the vertical coordinate of the neutral axis of the mid-ship
section ZNA is obtained at a prescribed curvature, the correspond-
ing reaction force of each stiffened panel over the cross section
can be easily obtained from its load–average strain curve accord-
Fig. 2 Finite element meshes. (Top) C4. (Bottom) S1 ing to its average strain derived from equation (13). Then, the
bending moment of the ship hull withstands under the prescribed
curvature given by
equilibrium condition, the reaction force over a hull cross section
F should be zero, which is indicated as M= 兺ⱍF z ⱍ
i i (15)

F= 兺F = 0
i (12) With the increase of the curvature of the mid-ship section, the
various stiffened panels are individually at different stages of elon-
where Fi stands for the reaction force over the cross section of the gation or shortening, and they may experience different local com-
stiffened panel i. Thus, the goal now is to determine the vertical pressive failure or tensile failure. The bending moment that the
coordinate of the neutral axis ZNA to satisfy equation (12). ship hull withstands will first increase with the increase of curva-
At first, an arbitrary ZNA is chosen for the initial position of the ture and then reach its maximum, namely the ultimate longitudinal
neutral axis. In accordance with the assumption that the normal strength Mult, at which some stiffened panels might collapse. After

188 SEPTEMBER 2008 JOURNAL OF SHIP RESEARCH


flange, and table are [06/452/06/45]s, [06/452/06/45]s, [06/452/06/
45]s, and [07/452/07/45]s, respectively. The geometries and dimen-
sions of stiffened panels at deck, side, and bilge are shown in Fig.
5 and Table 3, respectively. The ship hull is made of WR/polyester
GRP, and its material properties are shown in Table 4.
At the beginning of the progressive collapse analysis, the cross
section of the mid-ship is divided into a series of stiffened panels
and the resulting one-half of the cross section is shown in Fig. 4
where deck and side are composed of stiffened panels labeled 1
while bilge consists of panels labeled 2.
The next step is to develop the load–average strain curves for
stiffened panels 1 and 2 using the progressive failure nonlinear
finite element analysis described in section 2. The finite element
models of stiffened panels were developed using nine-node qua-
dratic degenerated shell elements. The finite element mesh of
stiffened panels is shown in Fig. 6 and the boundaries are simply
supported for loaded edges (along width) and symmetry along the
two unloaded sides (along length).
When the load–average strain curves of stiffened panels 1 and
2 are available as shown in Fig. 7, the moment–curvature rela-
tionship is established by imposing progressively increasing cur-
vatures of the cross section of the mid-ship.
At each prescribed curvature, the position of the instantaneous
neutral axis is computed by the trial-and-error process according
to the force equilibrium condition equation (12). Based on the
position of the instantaneous neutral axis, the average strain of
each stiffened panel is given by equation (13), and thus the cor-
responding Fi of each stiffened panel is subsequently obtained
from its load–average strain curve. The bending moment the ship
hull withstands at the prescribed curvature is then achieved by

Fig. 3 Graphs of end-shortening versus applied load. (Top) C4. (Bot-


tom) S1.

that, the capacity of the ship hull will decrease with the increase of
curvature because more and more stiffened panels have collapsed.

4. Numerical example

The progressive collapse analysis method described in the pre-


vious sections is exemplified by the prediction of the ultimate
longitudinal strength of an all-composite ship. The ship hull is a
longitudinally framed structure in which the length, breadth, and
depth are 50, 9, and 6 m, respectively. The cross sections of the
ship hull are formed by the hat stiffened composite panels, and
one-half of the cross section of the mid-ship is shown in Fig. 4.
The stacking sequences of plate, web, flange, and table of the
stiffened panels at deck and side are [04/45/04/45]s, [04/45/04/45]s,
[04/45/04/45]s, and [05/45/05/45]s, respectively. For the stiffened
panels at the bilge, the stacking sequences of the plate, web, Fig. 4 One-half of mid-ship cross section

SEPTEMBER 2008 JOURNAL OF SHIP RESEARCH 189


Fig. 5 Geometry of stiffened composite panels at deck, side, and bilge

Table 3 Dimensions of the stiffened composite panels of the composite ship (mm)

a b1 b2 b3 b4 b5 t1, t2, t4, t5 t3 d

Deck and side 1,000 300 62 50 57.3 18 10 12 60


Bilge 1,000 400 124 100 84.8 38 15 17 90

integrating the Fi of each stiffened panel over the cross section of failure of the stiffened panel 1 or 2 will suddenly occur. This
the mid-ship. With the increase of curvature of the cross section of phenomenon has been reported in the experiments in Starnes and
the mid-ship, the corresponding bending moment the ship hull Rouse (1981), Kong et al. (1998), and Smith (1990), which im-
withstands is obtained and finally the moment–curvature relation- plies an assumption that after each stiffened panel collapses, it has
ship of the ship hull is established. no strength and it sheds its entire load onto the other stiffened
After performing such a progressive collapse analysis, the re- panels. This is a simplification compared with the methods of
lationship between the vertical coordinate of the instantaneous Smith, which also account for the residual strength of the panels.
neutral axis and the curvature of the ship hull and the moment– However, while steel panels still have significant postcollapse
curvature relationship are shown in Figs. 8 and 9, respectively. strength, composite materials tend to exhibit a more brittle behav-
The ultimate longitudinal strength and the corresponding curva- ior and most fibers and matrixes have been fractured after col-
tures of the ship hull under sagging and hogging conditions are lapse, which justifies not considering the postcollapse strength.
shown in Table 5. Figure 8 shows that both for sagging and hogging conditions the
shift of vertical coordinate of the instantaneous neutral axis is in
5. Results and discussions general negligible, but the shift becomes significant when the ship
hull approaches collapse. A significant shift might be induced by
Figure 7 shows that the average tensile failure strain of stiffened stiffness reduction of some of the stiffened panels due to material
panel 1 or 2 is much higher than its average compressive failure failures when the ship hull approaches collapse. Also, it is noted
strain. This is because the buckling of stiffened panels under com- that with the increase of curvature the shift would decrease under
pression that occurred before collapse will greatly increase the the sagging condition but increase under the hogging condition.
stresses on the material and lead to the collapse of the stiffened Table 5 shows that the ultimate longitudinal strength and the
panels at low average strain while the collapse of a stiffened corresponding ultimate curvatures of the ship under sagging are
composite panel under tension usually does not occur until the both lower than the values under the hogging condition, which
average strain approaches its ultimate fiber strain. indicates that the ship under a sagging condition might be more
Figure 7 also shows that either compressive failure or tensile dangerous than it is under a hogging condition.

190 SEPTEMBER 2008 JOURNAL OF SHIP RESEARCH


Table 4 Material properties of the WR/polyester GRP

Property Symbol Value

Elastic modulus in 1 principal


material direction E1 15.8 GPa
Elastic modulus in 2 principal
material directions E2 15.8 GPa
Shear modulus in 1–2 and 1–3
principal material planes G12, G13 3.5 GPa
Shear modulus in 2–3 principal
material planes G23 0.35 GPa
Poisson’s ratios v12 0.13
Tensile strength in 1 principal
material direction XT 249 MPa
Compressive strength in 1
principal material direction XC 213 MPa
Tensile strength in 2 principal
material directions YT 249 MPa
Compressive strength in 2
principal material directions YC 213 MPa
Shear strength in 1–2 principal Fig. 7 Load–average strain curves of stiffened composite panels
material planes R 104 MPa
Shear strength in 2–3 principal
material planes S 23.5 MPa of the ship hull under sagging moments. However, the stiffened
Shear strength in 1–3 principal panels in the deck of most composite ships are commonly identical
material planes T 23.5 MPa or differ moderately in size or spacing. To avoid the catastrophic
accident, it is necessary to pay more attention to the evaluation of
the capacity of a ship hull made of composite materials under a
sagging moment.
A comparison between the progressive collapse analysis
method and a simplified method developed by Chen and Guedes
Soares (2007) is made herein. The ultimate longitudinal strength
in the simplified method is achieved by searching the equilibrium
failure points with use of the ultimate average strains of stiffened
composite panels. The detailed description of the method can be
found in Chen and Guedes Soares (2007) and is thus omitted here
for brevity.
The ship used in section 4 is utilized for this comparison. The
ultimate average strains of stiffened panels 1 and 2 forming the
hull cross section can be obtained from Fig. 7. With use of the
simplified method, the ultimate longitudinal strength and the cor-
responding curvatures of the ship hull under sagging and hogging
conditions are obtained as shown in Fig. 9.
Figure 9 shows that the predicted ultimate longitudinal strength
by the simplified method is 125 and 178 MNm for sagging and
hogging conditions, respectively. The predicted values are lower
by 0.8% and 3.8% than the corresponding values predicted by the
progressive collapse analysis method, which indicates that the
progressive collapse analysis method might provide a higher but
reasonable value for the ultimate longitudinal strength of ship
hulls than the simplified method, because the progressive collapse
analysis not only takes into account the particular sequence of the
collapse of each stiffened panel but also considers the shift of the
Fig. 6 Finite element mesh of stiffened composite panels instantaneous neutral axis of the mid-ship section due to the prop-
erty degradation of each stiffened panel before collapse.
Figure 9 shows that the reduced capacity of the ship after col-
lapse under a sagging condition is much more pronounced than 6. Conclusions
under a hogging condition and that the reduction is substantial.
This is because the deck of the ship is composed of a series of A progressive collapse analysis method is presented to predict
similar stiffened panels and that causes most of the stiffened pan- the ultimate longitudinal strength of ship hulls of composite ma-
els at the deck to lose their resistance capacities after the collapse terials, in which the behavior of each stiffened composite panel is

SEPTEMBER 2008 JOURNAL OF SHIP RESEARCH 191


Fig. 9 The moment–curvature relationship of ship hull. (Top) Sagging.
(Bottom) Hogging

Table 5 Ultimate longitudinal strength and the corresponding


Fig. 8 The relationship between the vertical coordinate of instanta- curvatures of the ship hull
neous neutral axis and curvature of cross section. (Top) Sagging. (Bot-
tom) Hogging Ultimate Longitudinal Strength Ultimate Curvature
(MNm) (1/m)

represented by its load–average strain curve, which is developed Sagging 126 2.95 × 10−3
by a progressive failure nonlinear finite element analysis. The Hogging 185 4.25 × 10−3
moment–curvature relationship of the ship hull is established by
imposing progressively increasing curvatures. The application of
the approach is exemplified by considering an all-composite ship
and the comparison with a simplified method shows that the pro- contract TNE3-CT-2003–506141. The work of the first author has
gressive collapse analysis method might provide a higher but rea- also been partially financed by Fundação para a Ciência e a Tec-
sonable value for the prediction of the ultimate longitudinal nologia, from the Portuguese Ministry of Science and Technology,
strength of ship hulls. under contract SFRH/BD/24310/2005. The financial support is
gratefully acknowledged.

Acknowledgments
References
This work has been performed within the project
MARSTRUCT—Network of Excellence on Marine Structures AHMAD, S., IRONS, B. M., AND ZIENKIEWICZ, O. C. 1970 Analysis of thick
(http://www.mar.ist.utl.pt/marstruct/) and has been partially and thin shell structures by curved finite elements, International Journal for
funded by the European Union through the Growth program under Numerical Methods in Engineering, 2, 419–451.

192 SEPTEMBER 2008 JOURNAL OF SHIP RESEARCH


BATHE, K. J. 1996 Finite Element Procedures, Prentice Hall, Englewood SLEIGHT, D. W., KNIGHT, Jr., N. F., AND WANG, J. T. 1997 Evaluation of a
Cliffs, NJ. progressive failure analysis methodology for laminated composite struc-
CALDWELL, J. B. 1965 Ultimate longitudinal strength, Transactions of tures, Proceedings, AIAA/ASME/ASCE/AHS/ASC 38th Structures, Struc-
Royal Institution of Naval Architects, 107, 411–430. tural Dynamics, and Materials Conference, April, Kissimmee, FL, 2257–
CHANG, T. Y., AND SWAWAMIPHAKDI, K. 1981 Large deformation analysis 2272.
of laminated shells by finite element method, Computers & Structures, 13, SMITH, C. S. 1977 Influence of local compressive failure on ultimate lon-
331–340. gitudinal strength of a ship’s hull, Proceedings, International Symposium on
CHEN, K. Y., KUTT, L. M., PIASZCZYK, C. M., AND BIENIEK, Practical Design in Shipbuilding, June, Tokyo, Japan, 73–79.
M. P. 1983 Ultimate strength of ship structures. Transactions of the So- SMITH, C. S. 1990 Design of Marine Structures in Composite Materials,
ciety of Naval Architects and Marine Engineers, 91, 149–168. Elsevier Science Publishers Ltd, London, England.
CHEN, N. Z., AND GUEDES SOARES, C. 2007 Longitudinal strength analysis STARNES, Jr., J. H., AND ROUSE, M. 1981 Postbuckling and Failure Char-
of ship hulls of composite materials under sagging moments, Composite acteristics of Selected Flat Rectangular Graphite-Epoxy Plates Loaded in
Structures, 77, 36–44. Compression, AIAA Paper, 81–0543, 423–434.
CHEN, N. Z., SUN, H. H., AND GUEDES SOARES, C. 2003 Reliability analysis TSAI, S. W., AND WU, E. M. 1971 A general theory for strength for aniso-
of a ship hull in composite material, Composite Structures, 62, 59–66.
tropic materials, Journal of Composite Materials, 5, 58–80.
ENGELSTAD, S. P., REDDY, J. N., AND KNIGHT, Jr., N. F. 1992 Postbuckling
response and failure prediction of graphite-epoxy plates loaded in compres- UEDA, Y., AND RASHED, S. M. H. 1984 The idealized structural unit method
and its application to deep girder structures, Computers & Structures, 18,
sion, AIAA Journal, 30, 2106–2113.
GORDO, J. M., AND GUEDES SOARES, C. 1996 Approximate methods to 227–293.
evaluate the hull girder collapse strength, Marine Structures, 9, 449–470. V A L S G A A R D , S., J O R G E N S E N , L., B O E , A. A., A N D T H O R K I L D S E N ,
GORDO, J. M., GUEDES SOARES, C., AND FAULKNER, D. 1996 Approximate H. 1991 Ultimate hull girder strength margins and present class require-
assessment of the ultimate longitudinal strength of the hull girder, JOURNAL ments, Proceedings, SNAME Symposium on Marine Structural Inspection,
OF SHIP RESEARCH, 40, 60–69. Maintenance and Monitoring, March, Arlington, VA, B.1–19.
KONG, C. W, LEE, I. C., KIM, C. G., AND HONG, C. S. 1998 Postbuckling YAO, T. 2003 Hull girder strength, Marine Structures, 16, 1–13.
and failure of stiffened composite panels under axial compression, Compos- YAO, T., ASTRUP, O. C., CARIDIS, P., CHEN, Y. N., CHO, S. R., DOW, R. S.,
ite Structures, 42, 13–21. NIHO, O., AND RIGO, P. 2000 Ultimate hull girder strength, Proceedings,
MOURITZ, A. P., GELLERT, E., BURCHILL, P., AND CHALLIS, K. 2001 Review 14th International Ship and Offshore Structures Congress (ISSC), October,
of advanced composite structures for naval ships and submarines, Composite Nagasaki, Japan, 321–391.
Structures, 53, 21–41. YUNUS, S. M., KOHNKE, P. C., AND SAIGAL, S. 1989 An efficient through-
SHAHID, I., AND CHANG, F. K. 1995 An accumulative damage model for thickness integration scheme in an unlimited layer doubly curved isopara-
tensile and shear failures of laminated composite plates, Journal of Com- metric composite shell element, International Journal for Numerical Meth-
posite Materials, 29, 926–981. ods in Engineering, 28, 2777–2793.

SEPTEMBER 2008 JOURNAL OF SHIP RESEARCH 193

You might also like