Composite Structures 62 (2003) 59–66

www.elsevier.com/locate/compstruct

Reliability analysis of a ship hull in composite material

Nian-Zhong Chen, Hai-Hong Sun, C. Guedes Soares *

Unit of Marine Technology and Engineering, Technical University of Lisbon, Instituto Superior T

ecnico, Av. Rovisco Pais,
1049-001 Lisboa, Portugal

Abstract
A simple and effective analytical method for ultimate longitudinal strength calculation and reliability analysis of a shipÕs hull in
composite materials is presented. The ultimate strength formula of composite stiffened panel is derived from composite column
theory. Deck and ship bottom structures are modelled as assemblies of stiffened panels. Then, progressive collapse analysis is applied
to calculate the longitudinal ultimate strength of ship hull in composite materials. The response-surface method and the first order
reliability method are combined to calculate the safety index and failure probability. Moreover, the sensitivity of each variable that
has effect on the reliability is discussed.
Ó 2003 Elsevier Ltd. All rights reserved.

Keywords: Composite; Stiffened panel; Ultimate strength; Progressive collapse analysis; Reliability analysis; Response-surface method; Sensitivity
analysis

1. Introduction flat, stiffened, carbon fibre composite panels under

compressive loading. Nonlinear finite element methods
In ships low weight is often desirable, but reduced were used to predict both the buckling and post-buck-
maintenance and corrosion resistance are also major ling response of the panels. A failure mechanism, in-
factors in design, which may lead to the selection of volving stiffener debonding, leading to eventual collapse
composite materials for the ship hulls. Different types of of the panel was identified and predicted using the finite
composite materials have been used [1,2] since high element method. These results may be useful to under-
performance composites used in yachts such as the ones stand the behaviour of fibre-glass panels under the same
in the AmericaÕs Cup races to the less sophisticated conditions.
applications of glass reinforced fibre-glass used in fishing Dow [8] presents the results of large and small-scale
boats [3] including the applications in naval ships [4]. experimental investigations into the collapse perfor-
Nowadays, more and more composite materials are mance of fibre-reinforced composites for marine struc-
being applied to high-speed ships both catamarans and tures. The author presents the results of strength tests
monohulls. carried out on stiffened panels, a variety of structural
Deck and ship bottom structures are assemblies of a connections and deck edge connections.
series of stiffened panels. Therefore, knowing the failure Structural failure of shipÕs hull girders due to extreme
capacity of stiffened panels is very important to assess bending moments is the most catastrophic failure. In
the ultimate strength of the ship hulls. Smith and Dow order to ensure safe design of a shipÕs hull, it is necessary
[5,6] investigated the compressive strength of hat stiff- to accurately evaluate the capacity of the hull girder
ened FRP panels, both theoretically and experimentally. under extreme loads, which lead to ultimate collapse.
They made recommendations regarding design proce- Caldwell [9] was the first who tried to evaluate the-
dures and safety factors on the design of FRP stiffened oretically the ultimate hull girder strength of a ship
panels. subjected to longitudinal bending. He introduced the
Stevens et al. [7] presented both theoretical and ex- plastic collapse moment considering the influence of
perimental results for the post-buckling behaviour of yielding of all structural members composing a shipÕs
hull, subjected to a bending moment.
*
Corresponding author. In CaldwellÕs method, the ultimate hull girder
E-mail address: [email protected] (C. Guedes Soares). strength is calculated without considering the strength
0263-8223/$ - see front matter Ó 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0263-8223(03)00084-9
60 N.-Z. Chen et al. / Composite Structures 62 (2003) 59–66

reduction in individual members after they have attained Lin et al. [16] use stochastic FEA and second order
their ultimate strength locally. This does not represent perturbation techniques to derive the strength of a
the real collapse behaviour of the structural members, laminate under in-plane loads based on first ply failure
but it is a conservative prediction. For this reason, it is using the Tsai–Wu criterion and buckling. The material
very important to take into consideration the strength properties, fibre angles and lamina thickness are treated
reduction (load shedding) of each structural member as random variables. SFEA and Monte Carlo simula-
when the collapse behaviour of a shipÕs hull is simulated. tion are used to develop failure probabilities using three
Some simplified methods have been developed to take different strength distributions (normal, lognormal and
into account of the strength reduction of structural Weibull). Randomness in lamina thickness is shown to
members after their ultimate strength as well as the time have the greatest effect on the failure probability of
lag in collapse of individual members. Smith [10] pro- angle ply laminates.
posed an approach in which the cross-section is divided Yushanov and Bogdanovich [17] gave state of the art
into small elements composed of a stiffener and attached overviews in reliability prediction methods for com-
plating, and the average stress–average strain relation- posite structures. The authors presented an analytic
ships of individual elements are derived before per- reliability prediction methodology and example appli-
forming a progressive collapse analysis, which accounts cations for laminates under in-plane loads, using anal-
for the contributions of all elements. ysis of rare passages of a state vector (stress, strain or
While SmithÕs method was based on finite element displacement) stochastic process beyond the stochastic
results for each stiffened plate element, Gordo et al. [11] limiting surface. The reliability of laminate is modelled
modelled this behaviour with simple analytical formulas as a series system of lamina reliabilities.
and comparison with experimental work showed the Jeong and Shenoi [18] presented a method using a
adequacy of the method [12]. simulation approach and Guedes Soares [19] has pro-
The research work that has been reported here on the posed different practical reliability approaches to assess
ultimate strength of ship hull girders was conducted on the reliability of composite structures.
steel structures. A similar work has not yet been per- In this paper a simple and effective analytical method
formed for structures of composite materials and this for longitudinal ultimate strength calculation and reli-
will be pursued in the present paper. ability analysis of composite ship hull in composite
Reliability analysis and design traditionally consider materials is presented. Based on composite column
ultimate limit state (ULS) to define a failure event. For theory, ultimate strength formula of composite stiffened
an ULS, the resistance or capability is represented by panel is derived. Then, progressive collapse analysis is
some measure of a structural strength, representing a applied to calculate the longitudinal ultimate strength of
maximum value of the structural resistance. Failure is composite ship hull girder, and the response-surface
said to occur when the predicted load or demand ex- method is combined with a first order reliability method
ceeds the predicted strength. The dominant strength to calculate the safety index and failure probability, and
failure modes are usually some form of collapse or the sensitivity of each variable that has effect on the
ductile overload. Proper inclusion of a strength predic- reliability is discussed.
tion in a structural reliability context requires the char-
acterisation and consideration of all possible strength
uncertainties. 2. Ultimate longitudinal strength of composite stiffened
In the case of components in composite material, panel
Sutherland and Guedes Soares [13] have reviewed the
state of the art in probabilistic models of composites. 2.1. Theoretical analysis
Ibnabdeljalil and Curtin [14] considered the strength of
composites and their size dependence with local load A stiffened panel of composite material is idealised as
sharing (LLS) using a Monte Carlo simulation model a beam-column, where w0 is the initial deflection of the
based on the 3-D lattice GreenÕs function technique. An stiffened panel, w is the plate deflection at any point, P is
analytical formulation was derived and demonstrated, axial compressive load, l is the length of the stiffened
mapping the reliability of a small composite structure panel, x is the coordinate in the direction of stiffened
under LLS to a large structure whose strength is gov- panel length, y is the coordinate in the direction per-
erned by its weakest bundle of fibres. Koksharov [15] pendicular to x and e1 is eccentricity of load, as shown in
used a characteristic volume with probabilistically Fig. 1.
characterised levels of damage in developing critical The initial deflection of the stiffened panel is assumed
distortion energy for the fracture strength of a unidi- to be
rectional composite. The approach allows inclusion of
composite structural reserve strength in reliability pre- px
w0 ¼ w0 max sin ð1Þ
dictions for both static strength and fatigue. l
 N.-Z. Chen et al. / Composite Structures 62 (2003) 59–66 61

and substituting it into Eq. (5) yields:

2   3
p2 e1
P4  e1 þ w0 max  w 0 max þ 8 P 5
rmax ¼ 1þ þ ð7Þ
A s s PE

Many composite materials including FRP behave like

brittle materials, showing no yield point. Therefore, the
Fig. 1. Composite stiffened panel under axial compressive loads. ultimate strength is reached when the maximum elastic
stress is reached:
2   3
p2 e1
where w0 max is the amplitude of initial deflection of  e þ w  w0 max þ 8 r
rULT 4 1 þ 5 ¼ Xc
1 0 max ULT
the stiffened panel. The plate deflection is given by þ
s s rE
w ¼ w0 þ w1 , where w1 is the bending deflection of the
stiffened panel. ð8Þ
The differential equation for the bending of com- or simply
posite beam-column is given by
r2ULT þ lrULT þ x ¼ 0 ð9Þ
d2 w1
Dv ¼
P ðw0 þ w1 þ e1 Þ ð2Þ where rULT is ultimate strength of the composite stiff-
dx2
ened panel and
where Dv is the flexural stiffness:
PE
X
N rE ¼ ð10aÞ
Dv ¼ ðEi0 A0i yi2 þ Ei0 Ii0 Þ A0
i¼1 Xc srE
x¼
2 ð10bÞ
Ei0
where is the YoungÕs modulus of the element i of the w0 max þ p 8e1
stiffened panel, A0i is its cross-sectional area, Ii0 is its rE ðs þ e1 þ w0 max Þ
moment of inertia and yi is the distance between the l¼ 2 ð10cÞ
centroid of the cross-section and the neutral axis of the w0 max þ p 8e1
cross-sectional area of the stiffened panel. Finally rULT is given by
The boundary conditions are pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

l þ l2
4x
x ¼ 0; l w1 ¼ 0 ð3Þ rULT ¼ ð10dÞ
2
Substituting Eqs. (1) and (3) into Eq. (2), leads to

 
w0 max px e1 kl kl 2.2. Failure mode of the composite laminate stiffened
w¼ sin þ cos kx

cos ð4Þ panels
1
a l cos kl2 2 2
where Theoretically, the primary models of overall failure
P p2 Dv 2 P for a stiffened panel subject to predominantly com-
a¼ ; PE ¼ 2 ; k ¼ pressive loads may be categorized into the following
PE l Dv
types [20]:
Therefore, the maximum stress rmax is given by
P Mmax 1. Overall collapse after overall buckling of the plating
rmax ¼ þ
A0 W and stiffeners as a unit.


P w0 max PE e1 p2 e1 P 2. Plate-induced failure by collapse at the corners of
¼ 0 1þ þ þ ð5Þ plating between stiffeners.
A s PE
P s 8 s PE
3. Plate-induced failure by collapse of a plate–stiffener
with combination at mid-span.
Dv 4. Stiffener-induced failure by local buckling of stiffener
W ¼
Ei0 y web.
where y is the vertical distance from the point of stiff- 5. Stiffener-induced failure by lateral–torsional buckling
ened panel to the neutral axis of the cross-section, A0 is of stiffener.
the cross-sectional area of the stiffened panel, s ¼
ðW =A0 Þ. Model 1 typically represents the collapse pattern
Adopting the approximation: when the stiffeners are relatively weak. The panel be-
haves as an ‘‘orthotropic plate’’. However, the stiffeners
PE P are designed stronger than the plate in practice, so this
1þ ð6Þ
PE
P PE mode is commonly avoided.
62 N.-Z. Chen et al. / Composite Structures 62 (2003) 59–66

Model 2 can occur in some cases when the panel is Table 2

predominantly subjected to biaxial compressive loads. The material properties of composite stiffened panel
Model 3 indicates a failure pattern in which the ultimate Wing Web Panel
strength is reached by column or beam-column type YoungÕs modulus E (MPa) 19.5 15.0 15.0
collapse of the plate–stiffener combination with the Tensile strength Xt (MPa) 241 238 238
Compressive strength Xc (MPa) 210 204 204
effective (reduced) plating.
Model 4 represents a failure pattern in which the
panel collapse by local buckling of stiffener web, while
Model 5 can occur when the ultimate strength is reached error 3.62% between the calculated result and experi-
by lateral–torsional buckling (or tripping) of stiffener. mental results is less than 5%.
The stiffeners in composite ship, which are usually of
‘‘hat’’ section, which are particularly effective in reduc-
ing the unsupported span of shell or deck panels and 3. Ultimate longitudinal strength of a composite ship
eliminating the possibility of lateral–torsional instability
under lateral or compressive load. So the model 5 is not The ultimate longitudinal strength of the hull girder
usual to occur in composite ship structures. is the value of the bending moment Mu at which the
In general, cross-stiffened panels should be designed flexural stiffness of the hull girder becomes zero. This
such that interframe collapse occurs before gross panel occurs when sufficient panels within a particular seg-
collapse because the latter involves a larger portion of ment of hull girder (the critical segment) have them-
structure and is likely to be more catastrophic. selves undergone collapse. During the overall collapse
As a result, the collapse modes of the stiffened panel the various panels are all at different stages of member
in composite under longitudinal compression are two collapse, and within each member the principal load
kinds: (1) Plate-induced failure by collapse of a plate– effect is the elongation or shortening which the hull
stiffener combination at mid-span; (2) Stiffener-induced girder bending moment imposes as a result of its cur-
failure by local buckling of stiffener web. In this mode, vature U. Because the collapse is constrained to occur
the plate has already been buckling before the failure, so between frames, the cross-section of the critical seg-
its resistance will be reduced. ment remains plane and the normal strain e due to hull
A reduction factor u [21] multiplied by the width of girder flexure varies linearly across the segment, in
panel b can be used to model this effect in a similar way accordance with the Euler–Navier beam bending the-
as is done with metal plates [22]. Ultimate strength of ory. That is
composite stiffened panel is the one whose ultimate e ¼ ZU ð11Þ
stress rULT is smaller than the other in the two failure
where Z is the distance from instantaneous elastic neu-
modes.
tral axis of the cross-section. According to the elastic
plastic bending theory, the equilibrium condition of
cross-section of the ship hull girder is
2.3. Example X
ri Ai ¼ 0 ð12Þ
Consider a composite stiffened panel, with an initial X
ri Ai Zi ¼ M ð13Þ
deflection of w0 ¼ 0:00013a, where a is the length of
composite stiffened plate. The geometric and material where Ai , ri and Zi represent the cross-sectional area of
properties of the composite stiffened panel are shown in the ith element, its average strain of the cross-section
Tables 1 and 2, and the geometry of the composite and the ordinate of its centroid of the cross-section re-
stiffened panel under longitudinal compression are spectively, and M represents the bending moment on the
shown in Fig. 2. cross-section of ship hull girder.
The calculated ultimate strength of the stiffened The y-coordinate of the instantaneous elastic neutral
plates is 29.428 MPa, while the experimental result of axis of the cross-section of the composite ship hull girder
the ultimate strength of the stiffened plates is 28.4 MPa Z can be derived from Eq. (4):
[1]. This shows that the method considered in this paper PN
Ei Ai Zi
is very accurate to calculate the ultimate strength of the Z ¼ Pi¼1 N ð14Þ
stiffened panel in composite materials since the relative i¼1 Ei Ai

Table 1
The geometric properties of composite stiffened panel
Name a b b2 b3 b4 bF t1 , t2 t3 t4 d
Size (mm) 3060 640 108 92 123 54 12.7 8.6 4 132
 N.-Z. Chen et al. / Composite Structures 62 (2003) 59–66 63

Fig. 2. The geometry and of the composite stiffened panel under longitudinal compression.

where Ei is the YoungÕs modulus of the cross-section of obtained by a linear regression method according to the
the ith element. form of Eq. (16).
Starting from ri and ei , one can determine Z from Eq. Then, the safety index b1 of the initial polynomial
(14) and then M from Eq. (13) after each stiffened panel limit state function will be obtained with a first order
collapse. Assuming that after each stiffened panel has reliability method. If the absolute value between b0 and
collapsed it has no residual strength, then it sheds its b1 is greater than a very small positive value f, it is
entire load onto the other stiffened panels. Step by step, necessary to redesign a series of test points. In this way,
Mu will be finally attained. a polynomial limit state function in the form of Eq. (16)
will be obtained until the absolute value between the two
adjacent safety indexes is less than f. The failure prob-
4. Reliability analysis ability Pf and safety index b of the ultimate longitudinal
strength of ship hull will be obtained finally.
The limit state function of the longitudinal ultimate Generally speaking, sensitivity analysis is an impor-
strength of a composite ship includes the model errors in tant part of structural reliability assessments. One ben-
predicting the ultimate longitudinal strength of the ship efit is the identification of the model parameters that
hull girder Cu , YoungÕs modulus of the material E, the have the most effect on the estimated reliability or safety
initial deflection of each stiffened panel w0 , the dis- index. The other benefit is being able to identify those
counting factor of the plates u, the compressive strength parameters that can be taken as fixed values and need
of material Xc , the model errors in predicting the lon- not be considered as random variables in reliability
gitudinal bending Cg and the longitudinal bending Mg . models.
The limit state function can be described as follows: If the limit state function is gðX1 ; X2 ; . . . ; Xn Þ ¼ 0, an
importance factor is defined by Ii :
g ¼ Cu Mu ðE; w0 ; u; Xc Þ
Cg Mg ð15Þ
 
where it can be seen that the ultimate longitudinal   
 og  
strength of the ship hull girder Mu is a function of E, w0 ,  r 
 
oXi P 
Xi 
u, Xc .  
Ii ¼  "  2 1=2 
# ð17Þ
The response-surface method is introduced since the  P

 n og  
 rX 
limit state function is complicated, and is not described  i¼1
oXi P  i 
by a simple analytical method. To construct a response
surface [23] a polynomial type function including where P  is the design point, Xi represents the ith vari-
squared terms to replace the original implicit limit state able, rXi is the standard deviation of Xi . The greater
function is adopted: importance factor Ii is, the more important Xi is.
X
n X
n
The sensitivity factor is defined by gi :
g¼pþ qi X i þ ri Xi2 ð16Þ
 
i¼1 i¼1  ob 

gi ¼  rXi  ð18Þ
where p, qi and ri represent the coefficients responding to orXi
Xi which is the ith variable.
Suppose an initial safety index b0 , and consider a where b is safety index. If a variable has a relatively large
series of E, w0 , u, Xc , Cg , Cu and Mg . A series of Mu will gi it must be treated in reliability analysis as a random
be obtained and then with the substitution of Mu , Cg and variable. On the contrary, a variable having a very small
Mg to Eq. (15), a series of corresponding g will be ob- gi can be regarded as a fixed variable in reliability
tained. An initial polynomial limit state function will be analysis.
64 N.-Z. Chen et al. / Composite Structures 62 (2003) 59–66

5. Example application draught is 2.3 m, load draught is 2.35 m, load dis-

placement is 400 t. The thickness of the plates are 4.1
Consider a high-speed ship built of fibre-reinforced mm at deck and side, and 7.2 mm at the bilge. The ge-
plastic. Its speed is 30 knots, it is longitudinally framed ometry of the stiffened panel in the mid-ship cross-sec-
structure, and the mid-ship cross-section is shown in tion is shown in Fig. 2. The geometry of the stiffened
Fig. 3, length overall is 50 m, length of waterline is 46 m, panels of deck and side are shown in Table 3, and Table
breadth moulded is 9 m, depth moulded is 6 m, designed 4 shows the characteristics of the stiffened panel of the
bilge. The dimensions a, b2 , b3 , b4 , bF , t1 , t2 , t3 , t4 and d
are shown in Fig. 2.
The ship is made of a composite material with
YoungÕs modulus E1 ¼ E2 ¼ E ¼ 1:8  1010 N/m2 , shear
modulus G ¼ 2:7  1010 N/m2 , PoissonÕs ratio l ¼ 0:26,
the YoungÕs modulus along the thickness of plate
Et ¼ E=10, the shear module along the thickness of plate
Gt ¼ G=10, the PoissonÕs ratio along the thickness of
plate lt ¼ 0:1. The ultimate tensile strength of the ma-
terial is Xt ¼ 2:08  108 N/m2 , and the ultimate com-
pressive strength of the material Xc ¼ 1:91  108 N/m2 .
The characteristics of the seven random variables Cu ,
E, w0 , u, Xc , Cg and Mg are shown in Table 5.
The failure probability and safety index of the high-
speed shipÕs failure probability and safety index in sag-
ging and hogging is shown in Table 6, where Pf is the
Fig. 3. The mid-ship cross-section. failure probability, b is the safety index.

Table 3
The geometry of the stiffened panel of deck and side
Name a b2 b3 b4 bF t1 , t2 t3 t4 d
Size (mm) 1000 62 50 57.3 18 4.1 3.5 2.1 60

Table 4
The geometry of the stiffened panel of the bilge
Name a b2 b3 b4 bF t1 , t2 t3 t4 d
Size (mm) 1000 124 100 84.8 38 7.2 6 3.1 90

Table 5
Characteristics of random variable
Number Name Mean Coefficient of variation Distribution
1 Cu 1 0.1 Normal distribution
2 E 1.8  1010 N/m2 0.06 Normal distribution
3 w0 0.0001 m 0.05 Normal distribution
4 u 0.6 0.15 Normal distribution
5 Xc 1.91  108 N/m2 0.06 Normal distribution
6 Cg 1 0.1 Normal distribution
7 Mg (sagging) 1.3534  107 N m 0.25 Extreme I distribution
8 Mg (hogging) 1.998  107 N m 0.25 Extreme I distribution

Table 6
Failure probability and safety index
Failure mode Pf b

4
Sagging DeckÕs compressive failure 6.0963  10 3.2343
Hogging BilgeÕs compressive failure 1.3595  10
9 5.9477
 N.-Z. Chen et al. / Composite Structures 62 (2003) 59–66 65

Table 7
Importance factor of each variable
Cu E w0 u Xc Cg Mg
Sagging 0.7189 0.3547 0.0006 0.3588 0.0011 0.4782 2.9262  10
6
Hogging 0.8279 0.2668 0.0019 0.3478 0.0123 0.3496 1.5165  10
6

Table 8
Sensitivity factor of each variable
Cu E w0 u Xc Cg Mg
Sagging 1.67 0.408 0 0.405 0 0.74 0
Hogging 4.23 0.42 0 0.645 0.006 0.73 0

The importance factor of each variable is shown in (3) The model errors in predicting the ultimate longitu-
Table 7, which shows that the importance factor of Cu is dinal strength of the ship hull girder Cu , the model
the maximum, second is the importance factor of Cg , errors in predicting of the longitudinal bending Cg ,
latter are the importance factors of u, E, Xc , w0 , while the reduction factor of plate u and YoungÕs modu-
the importance factor of Mg almost equals zero. lus of the material E must be dealt with as stochas-
The importance factor of Cu being very high indicates tic variables, while the compressive strength of
that Cu has a great effect on the safety index. After in- material Xc , the initial deflection of each stiffened
troducing Cg , the importance factor of Mg becomes al- panel w0 and the longitudinal bending Mg could
most equal to zero. be regarded as deterministic variables in reliability
Both the discounting factor of plate u and YoungÕs analysis.
modulus of the material E have also a great effect on
safety index.
The sensitivity factor of every variable is shown in Acknowledgement
Table 8, which indicates that the sensitivity factor of Cu
is the maximum, the sensitivity factors of Cg , u, E are The second author has been financed by Fundacßa ~o
fairly great, while the sensitivity factor of Xc , w0 , Mg are para a Ci^encia e a Tecnologia under the contract SFRH/
almost equal zero. This means that Cu , Cg , u, E must be BPD/5519/2001.
dealt with as stochastic variables and Xc , w0 , Mg could be
regarded as fixed variables.
References

6. Conclusion [1] Smith CS. Design of marine structures in composite materials.

London and New York: Elsevier Applied Science; 1990.
In this paper, a simple and effective method is pro- [2] Shenoi RA, Wellicome JF. In: Composite materials in maritime
posed for longitudinal ultimate strength calculation and structures. Practical considerations, vol. 2. Cambridge University
Press; 1993.
reliability analysis of a ship hull in composite materials. [3] Guedes Soares C, Roque R. Analysis of rule designed fishing
The method provides a relatively simple, quick and vessels in fibre reinforced plastics. Bull Assoc Tech Maritime et
reasonably accurate solution for the collapse of com- Aeronaut 1991;91:461–90.
posite stiffened panels, longitudinal ultimate strength [4] Mouritz AP, Gellert E, Burchill P, Challis K. Review of advanced
composite structures for naval ships and submarines. Compos
and reliability analysis of ship hull in composite mate-
Struct 2001;53:21–41.
rials. In addition, after performing a reliability analysis [5] Smith CS, Dow RS. Compressive strength of longitudinally
for a high-speed ship in composite materials, it was stiffened GRP panels. Compos Struct 1985;3:468–90.
concluded that [6] Smith CS, Dow RS. Interactive buckling effects in stiffened FRP
panels. Compos Struct 1987;4:1.122–37.
(1) It is most important to estimate precisely the ulti- [7] Stevens KA, Ricci R, Davies OA. Buckling and postbuckling of
composite structures. Composites 1995;28.
mate longitudinal strength of composite ship hull [8] Dow RS. Large scale FRP structural testing. In: International
since the model errors in predicting the ultimate lon- Conference Lightweight Materials in Naval Architecture, Royal
gitudinal strength of the ship hull girder Cu has a Institute of Naval Architects, 1996.
great effect on the safety index. [9] Caldwell JB. Ultimate longitudinal strength. Trans Roy Inst
Naval Architec 1965;107:411–30.
(2) Because both the reduction factor of plate u and
[10] Smith CS. Influence of local compressive failure on ultimate
YoungÕs modulus of the material E have a great ef- longitudinal strength of a shipÕs hull. In: Proceedings of Interna-
fect on safety index, it is also very important to esti- tional Symposium on Practical Design in Shipbuilding, Tokyo,
mate u and get the accurate statistical data of E. Japan 1977. p. 73–9.
66 N.-Z. Chen et al. / Composite Structures 62 (2003) 59–66

[11] Gordo JM, Guedes Soares C, Faulkner D. Approximate assess- [17] Yushanov SP, Bogdanovich SE. Analytical probabilistic model-
ment of the ultimate longitudinal strength of the hull girder. ling of initial failure and reliability of laminated composite
J Ship Res 1996;40(1):60–9. structures. Int J Solids Struct 1998;35(7–8):665–85.
[12] Gordo JM, Guedes Soares C. Approximate methods to evaluate [18] Jeong HK, Shenoi RA. Reliability analysis of mid-plane symmet-
the hull girder collapse strength. Mar Struct 1996;9(3–4): ric laminated plates using direct simulation method. Compos
449–70. Struct 1998;43:1–13.
[13] Sutherland LS, Guedes Soares C. Review of probabilistic models [19] Guedes Soares C. Reliability of components in composite
of the strength of composite materials. Reliab Eng Syst Safe materials. Reliab Eng Syst Safe 1997;55:171–7.
1997;56:183–96. [20] Smith CS, et al. Non-linear structural response. In: Proceedings
[14] Ibnabdeljalil M, Curtin WA. Strength and reliability of fibre- of the 7th International Ship Structures Congress, vol. 1, 1979.
reinforced composites: localized load-sharing an associated size p. II.2.1–20.
effects. Int J Solids Struct 1997;34(21):2649–68. [21] Zhenming W, Hai M. The calculation method of bearing capacity
[15] Koksharov II. An estimation of reliability of unidirectional com- of stiffened plates and shells of metal and composite materials.
posites by catastrophe theory. Mech Compos Mater 1996;32(4): Acta Mech Sinica 1985;17(3):253–8.
374–80. [22] Guedes Soares C. Design equations for ships plate elements under
[16] Lin SC, Kam TY, Chu KH. Evaluation of buckling and first-ply uniaxial compression. J Constr Steel Res 1992;22:99–114.
failure probabilities of composite laminates. Int J Solids Struct [23] Bucher CG, Bourgund U. A fast and efficient response surface
1998;35(13):1395–410. approach for reliability analysis. Struct Safe 1990;7:57–66.