Marine Structures 12 (1999) 1}19

Damage mechanics of top-hat sti!eners used

in FRP ship construction
H.J. Phillips , R.A. Shenoi
*, C.E. Moss
Hamworthy Marine Technology Ltd., Poole, UK

Department of Ship Science, University of Southampton, Southampton, Highxeld, Hants SO17 1BJ, UK
B.M.T. (Defence Services) Ltd., Bath, UK
Received 29 May 1997; received in revised form 14 January 1999; accepted 18 January 1999

Abstract
This paper is concerned with the assessment of damage tolerance of a top-hat sti!ener to
plate connection in FRP marine structures. The subject is addressed in two parallel schemes,
using stress-based and fracture-dependent criteria. Numerical modelling is used to determine
the internal load transfer characteristics and failure mechanisms in top-hat sti!eners under
typical loadings seen in practice. The "nite element models are benchmarked against published
test results, which include phenomena such as delaminations. The models are then extended to
include crack elements, which are employed to calculate the strain energy release rates in the
form of G values. The result from this modelling is compared against typical experimentally
derived data pertaining to G values for the materials in question. Finally, an attempt is made

to compare the results of the studies using the two approaches and to judge the overlap.  1999 Elsevier Science Ltd. All rights reserved.
Keywords: Fibre-reinforced plastic (FRP); Ship structures; Top-hat sti!eners; Finite element analysis;
Delamination; Strain energy release rates; Damage tolerance

1. Introduction
The sti!ness of large unsupported panels constructed of "bre-reinforced plastic
(FRP) materials is inherently low. Thus, it is necessary to sti!en such panels by a

* Corresponding author. Tel.: #1-44-1703-592316; fax: #1-44-1703-593299.

E-mail address: r.a.shenoi@ship.soton.ac.uk (R.A. Shenoi)
0951-8339/99/$ - see front matter  1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 5 1 - 8 3 3 9 ( 9 9 ) 0 0 0 0 3 - 9

H.J. Phillips et al. / Marine Structures 12 (1999) 1}19

Fig. 1. A typical top-hat sti!ener.

suitable method. The most prevalent form is the use of top-hat sti!eners. A typical
con"guration of the sti!ener is shown in Fig. 1.
The present fabrication method starts with the laying-up of the unsti!ened base
plate panel. Rigid foam cores are then laid on these panels in the locations where the
extra sti!ness is desired. Then, an adequate amount of "lleting resin is injected into the
recess between the foam and the plate panel and the required radius is scraped out.
Next, a resin-impregnated cloth, or overlaminate, is laid across the crown (or table),
down the web and around the "llet onto the base plate. A similar cloth run is carried
out for the opposite side. The process of overlaminating is repeated a number of times,
frequently with as many as 24 plies, to achieve the desired sti!ness. Occasionally,
a limited number of unidirectional plies may be applied to the crown to obtain extra
sti!ness. The reinforcement material for the web, #ange and overlaminate is E-Glass
woven roving and chopped strand mat; the resin matrix is isophthalic polyester resin.
Although top-hats are a critical part of the ship's structure, it has received surprisingly little attention in the open literature. Early reported work [1,2] was in connection with the development of the "rst GRP minehunter and dealt mainly with the
gross problems of qualifying design concepts with regard to a speci"c application.
This was extended to develop a more fundamental understanding of the top-hat
behaviour through theoretical modelling [3,4]. Such early modelling attempts were
necessarily restricted because of the relatively immature nature of the available "nite
element types. There was also some fundamental work on the use of novel "lleting
materials, with large strain to failure capability, and their application to marine joints

H.J. Phillips et al. / Marine Structures 12 (1999) 1}19

and top-hat sti!ening purposes [5]. E!ort was also now being directed towards
various failure scenarios. One of the principal cases involved overall panel instability
and sti!ener-plate interaction under compressive loading [4,6]. The second approach
focused on the transfer of load at the top-hat sti!ener to plate connection through the
orthogonal or out-of-plane direction under static loading conditions [7,8]. Guidance
on this subject in existing design regulations is minimal [9}11]. This is all the more
important because many ships have now been in operation for over 20 years and are
seeing sustained damage at the roots of the joints in the form of delaminations. It is
thus essential that a systematic study is carried out with regard to the localised
behaviour of a top-hat sti!ener to plate connection and its response under the
presence of cracks or delaminations.
The purpose of this paper is three-fold: (a) to understand the behaviour of a top-hat
sti!ener under static loading and the mechanisms of load transfer and failure; (b) to
study the in#uence of delamination cracks on strain energy release rates; and (c) to
propose a uni"ed approach to categorise damage tolerance levels in FRP structures in
the regions of top-hat sti!eners.

2. Strength-based assessment
2.1. Purpose of analysis
The main aim of this analysis is to identify how the load is transferred from an
in-plane direction in the base panel to an orthogonal direction in the plane of the web.
The transfer is achieved through the overlamination and "lleting material. It is
essential to categorise the stress states within the di!erent constituent elements of the
joint. These stresses can then be used to assess the likely causes of failure in the joints.
The three main possibilities are delamination in the overlaminate, cracking of the
"lleting resin and braking of the "bre plies in the overlaminate.
2.2. Physical characterisation of joint behaviour
Three load con"gurations were adopted for the test programme [12]; these were the
three-point bend, reverse bend and straight pull-o!, as shown schematically in Fig. 2.
Both the three-point and reverse bend tests aim to simulate gross panel deformation
and its e!ect on the top-hat connection. The pull-o! test is designed to simulate
inertial loadings on the sti!ener under impact and explosion loadings. In all cases the
load-de#ection plots were linear up until failure. The structural sti!nesses in the three
cases are listed in Table 1.
In case of the three-point bend test, specimens "rst showed damage at about
13.5 kN when the "llet-to-overlaminate interfaces failed. Subsequent loading caused
progressive through-thickness delamination in the inner region of the overlaminate
and the onset of cracking along the overlaminate-to-#ange joints from the cracks in
the "llet. Final failure occurred by #exural failure on the tension surface under the top
of the top-hat (see Fig. 3a).

H.J. Phillips et al. / Marine Structures 12 (1999) 1}19

Fig. 2. Loading con"gurations for the experiments: (a) three-point bend; (b) reverse bend; (c) pull-o!.

H.J. Phillips et al. / Marine Structures 12 (1999) 1}19

Table 1
Experimental failure loads and de#ections
Loading con"guration

Three-point bend
Reverse bend
Pull-o!

Sti!ness

(kN/mm)

Experimental

FEA

909
600
952

750
458
667

Fig. 3. Experimental failure modes: (a) three-point bend; (b) reverse bend; (c) pull-o!.

H.J. Phillips et al. / Marine Structures 12 (1999) 1}19

In the reverse bend tests, "rst signs of damage were observed at a load of 5 kN at
which point cracks appeared in the "llets on both sides. Final failure occurred at
a load of 14 kN due to tensile action on the inner surface at the centre of the #ange, as
shown in Fig. 3b.
In the pull-o! tests, cracks appeared at the #ange-to-"llet interface at a load of
about 5.5 kN. These cracks extended and joined together with increased load, until at
a load of 7 kN overlaminate was severed completely from the #ange on one side (see
Fig. 3c).
2.3. Features of the FE models
A series of models was generated in the ANSYS "nite element analysis package
using two dimensional solid elements (which possess three translational degrees of
freedom at each node); see Fig. 4 for a typical model. Each of the 12 layers in the
overlaminate was represented by one element through the thickness. The #ange plate
of the top-hat sti!ener has been represented by one element through the thickness.
Conditions of plain strain have been assumed throughout.
The loads applied to the structural model attempt to mimic those in the experimental investigation. For each of the three test con"gurations, stress distributions
have been computed (i) at the load at which initial damage was noted and (ii) at the
failure load of the sti!ener. The material properties used in the FE model generation
were obtained from previous work at Southampton [8] and from parallel work at
DERA Rosyth [12]; the relevant properties are given in Table 2.
2.4. Stiwness characterisation
The "rst step is to validate the FE models by comparing the model sti!ness with
that of the equivalent tested specimen. The FE model and experimental initial

Fig. 4. A typical "nite element model for strength calculations.

H.J. Phillips et al. / Marine Structures 12 (1999) 1}19

Table 2
Material properties used in the FE modelling
Material

Location

Property

Value

Polyester/woven
roving glass

Sti!ener, #ange
and overlaminate

Ex

13,060 MPa

Ey
nuxy
Ex
Ey
nuxy
Ex
Gxy
nuxy

7770 MPa
0.25
1500 MPa
1500 MPa
0.25
10\ MPa
10\ MPa
0.25

Urethane acrylate

Fillet

Core material

Table 3
Comparison of sti!nesses } experimental versus FEA
Three point bending
(3PB)

Reverse bending
(RB)

Straight pull o!
(PO)

FE
(N/mm)

Expt.
(N/mm)

FE
(N/mm)

Expt.
(N/mm)

FE
(N/mm)

Expt.
(N/mm)

731.2

696.8

713.0

384.6

930.0

1000

sti!nesses of the top-hat under each of the three loading con"gurations are shown in
Table 3. It is evident that for the three-point bend and the straight pull-o! tests there is
quite good correlation between the two sets of values. There is however a discrepancy
as far as the reverse bend results are concerned.
2.5. Stress patterns
The stress distributions of interest are the "llet principal stress, overlaminate
through-thickness and in-plane stresses, #ange plate through-thickness and in-plane
stresses. It is also necessary to compare the load transfer mechanisms predicted from
the FE models with some experimentally derived failure modes. Table 4 shows the
value and location of the maximum stress for each load level and load con"guration
for the top-hat sti!ener.
2.5.1. Three-point bending
The most signi"cant stress patterns for the top-hat at the experimental initial load
of 13.5 kN are shown in Fig. 5a for the overlaminate through-thickness stresses and
Fig. 5b for the #ange in-plane stresses. The magnitude of the "llet principal stress is the
greatest in the central region in the "llet as shown in Table 4 but is less than the

H.J. Phillips et al. / Marine Structures 12 (1999) 1}19

Fig. 5. Stress distributions in three-point bend con"guration: (a) overlaminate through-thickness stress;
(b) #ange in-plane stress.

Table 4
Locations and values of maximum stresses

H.J. Phillips et al. / Marine Structures 12 (1999) 1}19

10

H.J. Phillips et al. / Marine Structures 12 (1999) 1}19

ultimate value. Thus, the "llet is unlikely to fail at this load. The region under the
greatest in-plane stress in the #ange is in the inner central part below the core; this
however is unlikely to cause failure. The region of the #ange under the highest
through-thickness stress is in the outer central part as shown in Table 4. The region of
the overlaminate which is subject to both the highest in-plane and through-thickness
stresses is the outer region in the curved part above the "llet as shown in Fig. 5a.
Delaminations are likely to form here due to high through-thickness stresses.
The value of the maximum principal stress in the "llet at the sti!ener experimental
failure load of 16.5 kN is 18.09 MPa. The ultimate tensile strength (UTS) of the "llet
material in the literature [13] is quoted at about 26 MPa, so the "llet would remain
intact at this load. This corresponds to the failure mode in the experiments in which
the "llet itself did not crack. The initial damage was seen along the interface of the
"llet with the overlaminate. The through-thickness stress at the initial failure load of
13.5 kN, however, is greater than the quoted interlaminar tensile strength (ILTS) of
7 MPa for the woven roving/polyester [14]. Hence, the FE model predicts that
delaminations would occur near to the outer surface of the overlaminate at 13.5 kN
due to through-thickness stresses greater than the ILTS of the material. This exactly
matches the experimental "ndings. A similar match is obtained for the #ange failure
cause. At 16.5 kN the in-plane stress in the #ange is 208 MPa which is greater than the
UTS of the material. The FE model predicts that the #ange plate would fail in the
centre of the upper surface at a load of 16.5 kN, which again mirrors the experimental
"ndings.

2.5.2. Reverse bending

At a load of 5 kN, the "llet principal stress is 4.8 MPa which is much less than the
UTS of 26 MPa. The FE model, therefore, does not predict "llet failure at this load
level. The initial failure mode in the experiments, however, was that of "llet cracking.
The presence of voids within the "llet would cause higher stresses which could have
caused premature failure. This indicates that large voids may have been present in the
"llets prior to loading which opened out due to the nature of the load but did not
cause any further damage within the "llets. The experimental load/de#ection curve
showed no sudden loss of sti!ness and an FE model containing a void in the resin
exhibits an almost identical value of sti!ness as the model not containing voids. Thus,
it seems likely that the cracks in the "llet were due to the voids opening out under load
with no loss of top-hat sti!ness.
The in-plane stresses in the overlaminate are lower than the in-plane failure stress
but the through-thickness stresses in the overlaminate predicted by the FE model are
21 MPa along the interface of the overlaminate and the "llet. This is about three times
the ILTS so delaminations would be predicted in this location. No delaminations,
however, were visible in the experiments in this location. The high through-thickness
stresses may have caused a debond between the overlaminate and the "llet which in
turn caused the "llet crack. The FE model predicts maximum in-plane and throughthickness stresses in the #ange plate which are not high enough to cause failure at
a load of 5 kN. This is consistent with the experimental initial failure mode at 5 kN.

H.J. Phillips et al. / Marine Structures 12 (1999) 1}19

11

2.5.3. Straight pull-ow

The maximum values of stress for the "llet principal stress, overlaminate in-plane
and through-thickness and #ange in-plane and through-thickness stresses are given in
the lower two rows of Table 4. The "llet maximum principal stress at the sti!ener
experimental failure load of 7 kN is 7.8 MPa. This is much lower than the UTS of the
"llet material of 26 MPa. The FE model would not, therefore, predict "llet failure at
this load. This corresponds to the experimental failure mode in which no "llet cracks
were visible.
The maximum in-plane stresses in the overlaminate and in the #ange are less than
the UTS (in-plane) stress of 207 MPa at a load of 7 kN. Therefore, no failure is
predicted at this load from the FE model as a result of high in-plane stresses. The
maximum through-thickness stress of 2.8 MPa in the #ange is lower than 7 MPa
which is the ILTS. The maximum through-thickness stress in the overlaminate,
however, is higher than the ILTS. The FE model predicts delamination of the
overlaminate in the curved region close to the "llet due to high through-thickness
stresses.

3. Energy-based assessment
3.1. Fracture mechanics criteria used in the approach
Two dimensional linear elastic fracture mechanics (LEFM) models have been used
to calculate mode I and mode II stress intensity factors which, in turn, have been used
to evaluate strain energy release rates, G. The theoretical basis is outlined in Appendix
A. The load}de#ection characteristics of the top-hat sti!ener under the three modes of
loading discussed in the previous section are almost linear. Furthermore, in the
previous section, it was shown that the curved region of the overlaminate is one of the
most sensitive areas prone to delamination. The material in this location is linear
elastic up to failure. For this reason, there is no need to evaluate non-linear parameters such as the J-integral: only the strain energy release rate has been calculated in
each case.
3.2. Modelling details
The FE model used in the strength analyses has been adapted so as to include
a region containing cracks. The crack elements are six-noded triangular elements
with their mid-side nodes at the quarter point. The general scheme was similar to the
pattern shown in Fig. 4 except for the region containing the crack; this is shown
in Fig. 6.
Investigations showed that under certain conditions the two crack faces crossed
over each other, i.e. under a tensile load, the vertical displacement of the top crack face
was in fact less than the vertical displacement of the lower crack face. The problem of
crack faces overlapping has been discussed in the literature [15,16]. Four methods
exist which can be used to overcome this problem: (a) application of displacement

12

H.J. Phillips et al. / Marine Structures 12 (1999) 1}19

Fig. 6. Finite element detail in the region of cracks.

constraints on the crack face nodes, (b) application of nodal loads on the crack face, (c)
application of gap elements at the crack interface or (d) to assume that the overlapping
e!ect is negligible. In this particular case, a number of gap elements were inserted
along the crack face. The gap elements behave as linear springs in compression but in
tension their sti!ness drops to zero thus not inhibiting the crack face should it open. In
addition, the unloaded crack face is generated using nodes at the same location since
the gap element allows connection of two nodes which are initially coincident. A check
has been made to con"rm that the presence of the gap elements does not a!ect the
calculated values of the strain energy release rate. This has been done by comparing
the results from two models with and without the gap elements present. The two sets
of results are identical indicating that the presence of the gap elements has no e!ect on
the calculations.
3.3. Loads, material properties and boundary conditions
The material properties used in the "nite element model are given in Table 2.
Boundary conditions chosen were consistent with the strength analyses for the three
cases of three-point bend, reverse bend and straight pull-o!. The applied load in each
case is chosen as 10 kN. The signi"cance of this load is that it is below any

H.J. Phillips et al. / Marine Structures 12 (1999) 1}19

13

delamination damage which occurred in the three-point bend and reverse bend tests,
implying that any cracks simulating delaminations observed in the tests and appropriately inserted in the model should be stable at this load. Additionally, since the
strain energy release rates are proportional to the square of the applied load, it is
simple to interpolate values for di!erent load values.
3.4. Results
The study focused on cracks in the overlaminate region only, despite the fact that
the limited tests showed up some ultimate failures within the #ange laminate. This is
because the strength analysis showed the region of high stresses is in the curved
portion of the overlaminate, as indicated in Table 4. It was argued earlier that some of
the discrepancy in the static test results could have been due to fabrication variations.
Furthermore, parallel work on tee joints [17] shows that it is the curved region that is
crucial to the crack propagation tendency. Hence the focus of the fracture studies is
the overlaminate region.
Two types of crack variations for each of the three load cases were studied
} in#uence on crack depth and in#uence of crack length.
3.4.1. Crack depth
Fig. 7 shows the variation of G with crack depth. In Fig. 7a, pertaining to the
three-point bend, it can be noted that there is a peak value of G which corresponds to
a crack depth of 4 mm. Cracks which are deeper than 4 mm give rise to lower values of
G. It is anticipated that the reason why the value of G calculated for the crack at 2 mm
depth is lower than expected is due to the proximity of the crack to the surface. Cracks
close to the surface are more di$cult to model than those deeper within the overlaminate due to the limited area available to mesh with elements. This problem can be
avoided to a degree, by re"ning the mesh close to the surface. All the values of G are,
however, less than the critical value of 0.5 kJ/m [18]. This indicates that none of these
cracks under the three-point bend would propagate. The trend does, however, suggest
that under three-point bending, cracks which are deeper within the overlaminate are
less likely to propagate than those nearer the surface. Similar trends can be observed
for the reverse bend and straight pull-o! load cases as well, see Fig. 7b and 7c,
respectively. The strain energy release rates are of similar orders of magnitude as in the
three-point bend case.
3.4.2. Crack length
Fig. 8 shows the variation of G with crack length. Each crack is at a depth of 6 mm
from the outer surface of the overlaminate. From Fig. 8a for the three-point bend case,
it can be noted that the values of G increase at a steady rate as the crack length
increases. From the trends, it can be concluded that cracks greater than 38 mm in
length at a depth of 6 mm under these loading conditions are likely to propagate. For
the reverse bend condition, whose e!ect is shown in Fig. 8b, the G values are relatively
low indicating that this mode of loading is not as critical as the previous case. The
most severe condition in this respect is the straight pull-o!; the trends are indicated in

14

H.J. Phillips et al. / Marine Structures 12 (1999) 1}19

Fig. 7. Variation of G with crack depth: (a) three-point bend; (b) reverse bend; (c) pull-o!.

H.J. Phillips et al. / Marine Structures 12 (1999) 1}19

Fig. 8. Variation of G with crack length: (a) three-point bend; (b) reverse bend; (c) pull-o!.

15

16

H.J. Phillips et al. / Marine Structures 12 (1999) 1}19

Fig. 8c. The nature of the curve is similar to that in the three-point bend condition.
However, it can be seen that the fracture toughness value is exceeded at crack lengths
of about 30 mm.

4. Discussion
The results of the "nite element strength models are in good agreement with the
experimental results, particularly in the case of the three-point bending load. The results
have shown that the damage prone areas are (i) in the curved region of the overlaminate and (ii) in the central region of the #ange plate. This is indicated by the
presence of high through-thickness stresses in the curved region of the overlaminate
and also the presence of high in-plane stresses in the #ange.
The stress results have shown that the dominant failure modes of the top-hat
sti!eners under the three modes of loading include delamination in the overlaminate,
"llet-to-#ange interface cracking, "llet-to-overlaminate cracking, "llet cracking and
#exural failure in the #ange. In the case of damage tolerance calculations, it is the
initial failure mode which must be predicted since it is at this point when the
structure's load-bearing capabilities are gradually reducing. Thus, it is the initial
failure mode which is considered to be the most important in this case. In the case of
the three-point bending, the initial failure mode was in the form of delaminations in
the overlaminate. Damage in the #ange plate was also visible in both the three-point
bending and reverse bending cases but this was a "nal failure mode and is, therefore,
less signi"cant from a damage tolerance aspect. For these reasons, the delaminations
which have been modelled in the energy-based assessment which followed are those
within the curved part of the overlaminate.
The results of the strength and fracture approaches give similar results. For
example, the strength-based model of the three-point bending load shows that the
highest through-thickness stresses occur in the outer layers of the overlaminate. This
is shown in Fig. 5a. The energy-based approach shows that straight cracks which are
close to the outer surface of the overlaminate are more likely to propagate than those
which are deep. This is shown in Fig. 7a where greater values of G are obtained for
shallow cracks (not including the value at a crack depth of 2 mm). Also, the graph
showing the e!ect of crack length, shown in Fig. 8a shows that cracks which extend
into the curved part of the overlaminate are more likely to propagate than shorter
cracks which are contained within the #at portion of the overlaminate.

5. Conclusions
It has been shown that the delamination prone areas in top-hat sti!eners are
located in the curved region of the overlaminate close to the outer surface. The
delaminations are likely to be due to excessive through-thickness stresses. The damage
which occurs in the #ange is likely to be due to excessive in-plane stresses in the case of

H.J. Phillips et al. / Marine Structures 12 (1999) 1}19

17

three-point bending loads and due to excessive through-thickness stresses in the case
of reverse bending loads.
Calculation of fracture parameters with regard to delaminations in the top-hat
overlaminates, have shown that curved delaminations are most likely to propagate
under a straight pull-o! load. They are next likely to propagate under a three-point
bending load and are most stable under a reverse bending load. Also, delaminations
close to the surface are more likely to propagate than deep delaminations in the case
of all three loading scenarios.
Results from the strength-based approach and the energy-based approach are
comparable and similar trends have been found.

Acknowledgements
The work presented in this paper was funded by the EPSRC/MoD. We are grateful
to Lt. Cdr. Mark Gray (MoD), Prof. John Sumpter, Richard Court, David Elliott,
Philip Lay, Andrew Swift and Richard Trask (DRA) for their helpful discussions
during the project.

Appendix A. Fracture mechanics criteria for damage modelling

A.1. Elastic stress xeld approach
Irwin [19] developed the stress intensity approach from linear elastic theory. In the
region of the crack tip, the stress intensity factor, K, determines the magnitude of the
elastic stresses. The value of K, shown in Eq. (A.1) depends upon the magnitude of the
applied stress, p, the length of the crack, 2a, and a parameter which depends upon the
crack and specimen geometry, f (a/=) where = is the specimen width:
K"p(pa f



a
.
=

(A.1)

Irwin proved that the achievement of a critical stress intensity factor, K , is exactly
!
equivalent to the Gri$th}Irwin balance approach. This requires that the achievement
of a stored elastic strain is equal to the critical strain energy release rate, G [20]. For
!
tensile loading, the relationship between K and G is given in Eq. (A.2) for plane
!
!
stress,
K
G " ! plane stress.
!
E

(A.2)

All stress systems in the vicinity of the crack may be derived from three modes of
loading: (a) mode I which is the opening mode, (b) mode II which is the sliding mode
and (c) mode III which is the tearing mode.

18

H.J. Phillips et al. / Marine Structures 12 (1999) 1}19

A.2. Energy balance approach

The mode I strain energy release rate, G , can be written in terms of the mode
'
I stress intensity factor, K , from Eq. (A.2) and with f (a/=) in Eq. (A.1) being equal to
'
unity (i.e. an in"nitely long plate) as
K
G " ' plane stress.
'
E

(A.3)

The strain energy release rate, G, can be considered to be the amount of energy which
is available for crack extension and can be written in terms of the three stress intensity
factors for mixed-mode behaviour:
(K#K )(i#1) K
''
G" '
# ''',
8k
2k

(A.4)

where K is the mode I stress intensity factor, K is the mode II stress intensity factor,
'
''
K is the mode III stress intensity factor, k is the material shear modulus, i is the
'''
conversion factor between conditions of plane strain and plane stress, and is equal to
(3!4t) for plane strain conditions and t is the material Poisson's ratio.
For the case where only modes I and II are applicable, mode III is assumed to give
a negligible contribution to the strain energy release rate and hence the second term in
Eq. (A.4) can be neglected.

References
[1] Dixon RH, Ramsey BW, Usher PJ. Design and build of the GRP hull of HMS Wilton. Proceeding of
the Symposium on GRP Ship Construction, RINA, London, 1972;1}32.
[2] Smith CS. Structural problems in the design of GRP Ships, Proceedings of the Symposium on GRP
Ship Construction, RINA, London, 1972;33}56.
[3] Clark MA. Stress analysis of a frame-skin connection in the MCMV. NCRE Report No.
NCRE/L8/76, March 1976.
[4] Smith CS. Design of marine structures in composite materials. Amsterdam: Elsevier, 1990.
[5] Green AK, Bowyer WH. The development of improved attachment methods for sti!ening frames on
large GRP panels. Composites, 1981;12:49}58.
[6] Smith CS, Dow RS. Interactive buckling e!ects in sti!ened FRP panels. In: Marshall IH (Ed.),
Composite Structures Amsterdam: Elsevier, 1987;122}134.
[7] Shenoi RA, Hawkins GL. An investigation into the performance characteristics of top-hat sti!ener to
shell plating joints. Compos. Struct 1995;30:109}121.
[8] Dodkins AR, Shenoi RA, Hawkins GL. Design of joints and attachments of FRP Ships' structures.
Marine Struct 1994;7:365}98.
[9] Rules for Yachts and Small Craft, Lloyds Register of Shipping, London, 1983.
[10] Rules for Building and Classing Reinforced Plastic Vessels, American Bureau of Shipping, New York,
1978.
[11] Rules for Classi"cation of High Speed and Light Craft, Det Norske Veritas, 1991.
[12] Elliott DM. Mechanical testing of composite joints } Interim Report. DRA/AW/AWS/TR94212,
April 1994.
[13] Shenoi RA, Hawkins GL. In#uence of material and geometry variations on the behaviour of bonded
Tee connections in FRP Ships. Composites, 1992;23:335}345.

H.J. Phillips et al. / Marine Structures 12 (1999) 1}19

19

[14] Bird J, Allan RC. The determination of the interlaminar strength of ship type laminates. in:
Proceedings of the 7th International Conference Experimental Stress Analysis, Haifa, 1982:91}104.
[15] Pavier MJ, Clark MP. A specialised composite plate element for problems of delamination buckling
and growth. Composite Struct 1996;34:43}53.
[16] Tian Z, Swanson SR. E!ect of delamination face overlapping on strain energy release rate calculations. Composite Struct 1992;21:195}204.
[17] Phillips HJ. Assessment of damage tolerance levels in FRP Ships' structures. PhD thesis, University of
Southampton, March 1997.
[18] Court R. Fracture toughness of woven glass reinforced composites. Interim Report
DRA/SMC/CR943111, December 1994.
[19] Irwin GR, Kies JA. Critical energy rate analysis of fracture strength. Welding J Res Suppl
1954;193s}198s.
[20] Ewalds HL, Wanhill RJH. Fracture mechanics. Publs. Arnold & DUM, 1986. ISBN 0 7131 3515 8.