An investigation into the performance characteristics of top-hat stiffener to shell plating joints

Composite srructures 30 ( 1995) 109- 12 1 0 1994 Elsevier Science Limited Printed in Great Britain. All rights reserved 0263~8223/95/S9.50 0263-8223(94)00032-S

ELSEVIER

An investigation into the performance characteristics of top-hat stiffener to shell plating joints R. A. Shenoi & G. L. Hawkins Department of Ship Science, University of Southampton, Highjield, Southampton,

UK, SO9 SNH

This paper is concerned with the design of top-hat stiffener to shell plating joints such as those typically found in fibre-reinforced plastic (FFZP) ships and boats. The background to the problem is presented, as characterised by a review of past work and current design/production procedures. From these considerattons influential geometric vanables are defined, and a ‘new’ design of stiffener joint is developed. Finite element modelling is used to compare the behaviour of this joint with typical current practice. Load transfer paths and failure mechanisms are determined. Finally the results of a study are presented in which the key design parameters are systematically varied to highlight their influence.

applied at the top of the table (or crown) to obtain extra stiffness. This is a time-consuming and expensive process. Figure l(A) illustrates this type of stiffener. An alternative form of building up the stiffness, which requires study, is owed to the increasing availability of pre-formed sections and the ease of producing large and void-free fillets afforded by the mechanised gun. This is shown in Fig. l(B). The proposed method would involve placing the pre-formed section at the right location and being bonded to the base panel by the injecting of filleting resin (as in the case of the foam former). This would then be capped by two or three layers of woven roving overlaminate. A key characteristic of these out-of-plane joints is that, because of a lack of continuity of reinforcing fibres across the joint, they are susceptible to fail by peel or delamination well before the ultimate in-plane material stress is reached. Furthermore, their dependence on interlaminar properties make them somewhat sensitive to material imperfections such as voids and to minute changes in geometry in the laminate. Because of such sensitivity in structural performance and the weight/cost implications involved, it is important to ensure that the design and production of such joints is carefully carried out. The purpose of this paper is to study the problem of top-hat joints and to identify key variables that control and govern the transfer of load from

1 INTRODUCTION A major constraint to overcome in the design and construction of large fibre-reinforced plastic (FRP) structures is the relatively low modulus of the materials. Specific stiffness of FRP is a third to a quarter of the value in metals. This feature leads to the need for stiffening large, unsupported spans of plating panels by appropriate mechanisms. The most prevalent form is the use of top-hat stiffeners. The current method used to fabricate such stiffened structures starts with the laying-up of the unstiffened panel. Rigid foam cores are then laid on these panels in the appropriate locations, where stiffness is desired. Next, an adequate amount of filleting resin is injected in the recess between the foam and the plating panel, and the required radius is scraped out by a curved spatula. More recently the fillet is applied and radiused in one pass using a filleting gun, increasing both the speed of operation and the quality of the fillet. The process of overlaminating now takes place: the resin-impregnated cloth is laid across the table, down the web and around the fillet onto the base plate panel. A similar cloth run is carried out for the opposite side. The process of overlaminating is repeated a number of times as necessary, frequently as many as 14 plies, to obtain the required stiffness. Occasionally, a limited number of unidirectional plies may be 109

110

R. A. Shenoi, G. L. Hawkins

the panel to the stiffener and vice versa. This is done first by gaining an understanding of the internal mechanisms of load transfer, and then determining the impact of variations in the geometry of the joint.

2 CHARACTERISATION

Stiiener laminateddirectly

onto cured platii

OF THE PROBLEM A: Cunent Designof Stiffaner.

The large differential in stiffness between the tophat section and the shell plating results in the joint between them being a highly loaded region. The forms of loading that are applied to these quasi two-dimensional structures in a marine context are limited to in-plane loading (compressive or tensile) such as that which results from longitudinal bending of the hull girder, and out-of-plane loading resulting from hydrostatic pressure, berthing and docking, tank pressure testing, and explosion trials. Previous workle3 has shown that the joint region is strong relative to the rest of the top-hat/plate section during typical in-plane loading, but is the area of weakness under out-of-plane loading. Hence the work outlined here has concentrated on the out-of-plane loading scenario. Out-of-plane loads can be applied to the tophat section in two forms, those which cause the overlaminate to be either in compression, or in tension. These conditions can be developed by several different loading methods, using either tensile or compressive forces. However the application of tensile forces has received the most attention since the primary problem has been joint debonding, not stiffener web buckling.4 The two loading conditions derived by applying tensile loading are illustrated in Fig. l(C). The centreclamp loading (overlaminate in compression) has been commonly used as a test method5v6 as it results in the lowest failure load. It is developed by clamping the shell plating in the centre of the tophat, and applying a tensile load on the table of the top-hat. However, it does not result in premature delaminations occurring in the overlaminate, and is unlikely to occur under tension in a ship’s structure since the width of unsupported plate between stiffeners is generally very much greater than the width of the stiffeners themselves. Two-clamp loading (overlaminate in tension) is that which would typically occur when a tank is subjected to test, or when a shock bubble implodes resulting in a negative pressure on the shell plating. This is developed by clamping the shell plating on either side of the top-hat, and applying a tensile load to the table of the top-hat. This loading method also

U-sectionore-formed

overtaminate

bp

of pa-determined size

6: New Designof Stiffener.

Centre Clamp.

Two Clamp.

C: BoundaryConditionsApptiedto Modek.

Fig. 1.

Top-hat stiffener models.

results in the premature delaminations seen in practice under these loading conditions.7 To allow comparison between this and previous work, both centre- and two-clamp loading have been considered here. The geometric variables associated with this joint are shown in Fig. 2. For the current method of construction the backfill angle is controlled by the undercut in the foam former. The gap does not exist as such, since the web continues around the fillet radius onto the shell plating to form the overlaminate. Hence the thickness of the overlaminate is controlled by the thickness of the web, which is dependent on global stiffener requirements. Thus the only variables not predetermined are the fillet radius and the backfill angle, allowing very little optimisation to be done. For the alternative construction method proposed above in Fig. l(B), all the geometric variables are free from global constraints and thus can be optimised for production and performance. A review of current

Performance characteristicsof a top-hat stiffener

\GAp

OVERLAMINATE

THICKNESS

Fig. 2.

Key top-hat joint design variables.

practice in the marine industry was undertaken to determine the emphasis currently given to these geometric variables, and to assess the requirements for a detailed study. One of the earliest approaches to FRP structure design was outlined in Gibbs and COX.~This gives recommended arrangements of various joints with simple design examples. Section moduli and moments of inertia are tabulated for different geometries of top-hat stiffeners. The dimensions of the boundary angles and overlaminates, it is stated, ‘should be minimum compatible with strength requirements’. However, no specific procedures concerning joint design are elaborated. Early work in the UK centred around the design of FRP minehunters and this formed the impetus to the drawing-up of naval engineering standardss9 Early designs specified the use of bolts through both the base plate and the overlaminate to double-up on the bonded connection, thus ensuring that total peeling of the top-hat from the base plate did not occur. This practice was superseded by the use of bonded connections alone, with the bonding being achieved by the use of flexible, urethane acrylate resin for the fillets.” The rules for naval ships are based primarily on extensive experimental work. The standards prescribe minimum limits to various scantlings. Overlaminate thickness, for example, is specified to be at least half (and preferably two-thirds) the thickness of the thinnest member at the joint. Flange overlap dimensions, the lay-up and stacking sequence are also specified. With regard to non-naval craft, design guidance is sought primarily from classification society rules. Lloyd’s Register of Shipping ruleslo state

111

that, for top-hat stiffeners, the ‘width of the flange connection to the plate laminate is to be 25 mm + 12 mm per 600 g per sq. m. of reinforcement in the stiffener webs, or 50 mm, whichever is the greater’. American Bureau of Shipping rules” state that ‘the minimum overlap on the plating should be 20% of the web depth or 50 mm, whichever is greater’. There is also a minimum thickness requirement for the overlaminate. Det Norske Veritas rules12 specify that first principles based calculation procedures should be adopted to determine section moduli of top-hat stiffeners and scanthngs of plate laminates. However, no explicit guidance is given with regard to the requirements, standards or design allowables. This lack of specificity is further evident from a recent survey of current practice13 where the design procedures for top-hat are merely updates on the Gibbs and Cox practice. Clearly there is little emphasis given to the optimisation of the geometric variables associated with these joints. The main thrust of current practice seems to be one of aiming to make the top-hat stiffener to base plate connection as stiff as possible. Furthermore, there is a mistaken belief that increased stiffness also corresponds to increased strength of the joint. This fallacy has been exposed for other similar bonded connections, e.g. tee-joints. I4 The procedure adopted for the work outlined in this paper follows the tee-joint approach which focused on an explicit identification of design variables and production processes influencing joint behaviour. The present study has centred around determining the effect of the key design variables thought to influence the behaviour of the joint, illustrated in Fig. 2. Because of the lack of analytical approaches, giving exact solutions, recourse was made to numerical techniques for modelling the joint behaviour.

3 FINITE

ELEMENT

MODELLING

All the modelling was done using the ANSYS suite of programs. l5 This offers three elements with composite capabilities, two shell, and one three-dimensional solid. The shell elements are defined by eight nodes on the plane at mid-thickness, allowing elements to be linked end to end but not on top of each other. The geometry of the top-hat section means that some stacking of elements is required so the three-dimensional solid element is used throughout. This element is

112

R. A. Shenoi, G. L. Hawkins

defined by eight nodes, each having three degrees of freedom (translations in the nodal x, y, and z directions), and has large deflection and stress stiffening capabilities. Several preliminary analyses of the tee-joint problemI were conducted to assess the effects of incorporating these capabilities, and also the effect of varying the mesh density. This indicated that the large deflection option is required for the more flexible designs, and that the mesh density should be as refined as possible, particularly in any region where the laminate is curved or is subjected to high through-thickness loads. The area of the overlaminate where it passes around the fillet is an example of this. Memory and wavefront considerations apply an upper limit to this but nevertheless a reasonably high level of definition was achieved (typically 600 nodes, and 300 elements, with 1500 degrees of freedom). Models were developed with a consistent mesh density so that any mesh density effects were eliminated from the parametric results. These were run using the large deflection option. The material properties used were derived from manufacturers’ data (where this was available) and from experimental results. The properties are listed in Table 1. Several features should be noted about these materials. Firstly, there is a considerable differential between the throughthickness and in-plane strengths of the laminates. Indeed, the through-thickness strength is considerably less than the ultimate tensile stress (UTS) of the matrix material due to the stress concentration Table 1. Material properties used in finite element analysis Material ($a)

Polyester CR1 152” CR1200”

’

UTS (CL?&) (MPa)

Failure strain (“A) 6 100 27

3.2 0.5 0.7

-

-

58 26 32

14.68 13.06 -

0.123 0.139 -

3.09 3.09 -

207 207 12.2

6.89 -

0.13 -

3.45 -

11.2

CR1200/WR Warp Weft

6.375 3.926

-

-

183.3 188.9

2.8 4.8

CR1 200/CSM In-plane

3.023

-

-

110.4

3.6

Polyester/WR Warp Weft Interlaminar Polyester/CSM In-plane Interlaminar

“CR1 152: urethane acrylate resin. %R1200: polyester/urethane acrylate mix.

1.4 1.4

effects caused by the presence of the fibres. Thus these materials are prone to failure in the throughthickness direction at relatively low stress levels. Secondly, the properties of the fillet material are nonlinear and incorporate very large plastic deformation before failure, making it necessary to apply the load incrementally to follow this plasticity. Figure l(C) shows the boundary conditions applied to the models, which represents centreclamp and two-clamp loading. The development of these test methods is discussed above, and both are included in this study to allow validation of the finite element models with previous work, to allow a comparison to be made between the different loading methods, and thus to determine the most appropriate form of loading to produce the failure mechanisms seen in full-scale explosion trials.7

4 DISCUSSION Initial results are taken from a study to assess the relative performance of the ‘new’ design of tophat when compared to the current method of construction. For validating the finite element modelling the load/deflection characteristic of the model of the current design was compared to existing experimental data.6 As there is no experimental data available to validate the model of the new joint design, a consistent model definition has been maintained from the current to the new design to limit any variations. Figure 3 shows load/deflection curves (measured at the point of load application) for the current design, both experimental and theoretical, and for a nominal new design with a fillet radius of 75 mm, a gap of 20 mm, a backfill angle of 45”, and an overlaminate thickness of 2 layers of woven rovings, all for centre-clamp loading. This shows close correlation between experimental and theoretical results, and also shows that the overall stiffness of the tophat section is little affected by the design of the joint. This last result is explained in Fig. 4(A) and (B). This shows the deflection distribution through the model of the new design, for centre-clamp and two-clamp loading respectively. The location of each point is shown in Fig. 4(C). For centre-clamp loading the displacement of the base panel is very similar in magnitude to the overall displacement at the flange of the top-hat, and for two-clamp loading it is a significant proportion of the overall

Pe$ormance characteristicsof a top-hat stiffener

0

0

5

10

15

20 had

Fig. 3.

Comparison

25

30

35

40

45

113

50

(KN)

of load/deflection

displacement. This indicates that the deflection of the base panel has the most influence on the overall deflection. Since the construction of the base panel is similar for all models, their deflections are also similar. This important result shows that the function of the stiffener, that is to stiffen the base panel, is a function of the stiffener scantlings, and not the design of the joint between the stiffener and the base panel. This means the design of this joint can be refined to improve its performance without impairing the overall performance of the stiffener itself. However, the internal behaviour of the joints varies dramatically, as shown by the stress plots shown in Figs 5 and 6. Figure 5 shows the inplane stress in the overlaminate, the throughthickness stress in the overlaminate, and the maximum principal stress in the fillet for the current design of joint subjected to two-clamp loading. Figure 6 shows similar results for the nominal new design at the same load level. In the current design the dominant in-plane stress in the overlaminate (Fig. 5(A)) is tensile of magnitude 197 MPa (very close to failure) in the root of the radius, with the maximum compressive stress of - 141 MPa occurring on the inner face of the web due to the bending applied to the web by the unaligned boundary conditions (this can be seen in Fig. l(C)). Clearly the overlaminate will fail in tension before compression. With the new design, however, the dominant in-plane stress in the overlaminate (Fig. 6(A)) is now compressive ( - 152 MPa) on the inner face of the web whilst the maximum in-plane tensile stress is reduced to 117 MPa. This occurs on the outer face of the web, not

curves (centre-clamp).

in the overlaminate, and is thus not affected by the laminate curvature which increases stress levels. Hence in the nominal new design the web will fail in compression at a higher load than the current design. Optimisation of the geometric variables should ensure that the web fails in tension and compression at the same load level. Figure 5(B) shows the high (21 MPa) throughthickness stress in the overlaminate of the current design occurring in the centre of the overlaminate in the root of the radius. This value is 2.5 times the failure stress (from Table 1) and corresponds to the position of first failure seen in trials. In the new design (Fig. 6(B)) the through-thickness stress in the overlaminate is reduced to above 5 MPa with the maximum of 13 MPa occurring in the web. Thus the initial failure will now occur in the web itself (at a 50% higher load than the current design) and the problem of premature delamination of the overlaminate is removed. The maximum principal stress in the fillet of the new design (Fig. 6(C)) is 24 MPa compared to 16 MPa for the current configuration (Fig. 5(C)). The location is similar in both cases, being against the overiaminate adjacent to the positions of maximum through-thickness stress. The fillet in the new design is thin in this location, contributing to the higher stress levels, but these are still less than the UTS of the fillet material. Note also that the stress level on the end of the web is less than 3 MPa, indicating that this discontinuity does not result in high stress concentrations. Overall the results in Figs 5 and 6 indicate that the current design will fail by premature delamination of the fillet at a very low load (this corre-

R. A. Shenoi, G. L. Hawkins

114 25

20

_

15

-.

r

Top of Web

--- Centre of Flct --f3ass of Fiiat

5

0 0

A:

10

Deflection

20

30

Distt%Z%%

40

50

Curves

60

70

(Centre

Clamp)

25

-. Top of Web --Top of F&A --Bass of F&t

B: Deflection

Dish%%%

Curves

(Two

Clamp)

Overall Defl Top of Web

C: Locations of Above Deflections. Fig. 4.

Deflection distribution

curves in centre-clamp

sponds to observed failure during experimentation) whilst the nominal new design will fail by fillet failure at a load level approximately three times higher. Optimisation of this joint should improve this performance to utilise better all the materials in the joint.

and two-clamp loading.

Once the functionality of the new design had been ascertained a parametric analysis was conducted to study the effects of changing the four main geometric variables shown in Fig. 2. The results of this analysis are shown in Figs 7-9. These show the overall displacement, maximum

Performance characteristics ofa top-hat stifiener

A: In-plane

Stress in Overlaminate.

B: Through-Thickness

Stress in Overlaminate.

C: Principal Stress in Fillet. Fig. 5.

Stress distributions in boundary angle overlaminate and fillet - base model.

115

116

R. A. Shenoi, G. L. Hawkins

A: In-plane

Stress

in Overlaminate.

B: Through-Thickness

C: Principal Fig. 6.

Stress distributions

Stress

in Fillet.

in boundary angle overlaminate

and fillet - ‘new’ model.

Peeormance

characteristics of a top-hat stiffener

117

&--/ 1

2

Effect

81 I Effect

3

2

3

Effect

5

6

cwwkxnkat~

Overlaminate

J

of

4

4

3

6

ovvimlinat.

Overlaminate

7 8 thic*ne

6

Overlaminate

of

+1

5

4

O”edrzfnim.ate

of

on

7

9

Stress

B thickno

on

its

7

II

ws*na on its

x)

11

12

in Fillet

9

lo

In-plane

9

T-T

10

n

in Fillet

12

Stress

n

2

Effect

3

of

4

5

-2

G-admh¶ta

Overlu-ninde

7

8 rmdva

on

its

9

h-plme

10

11

12

Stress

12

Stress

Centre Clamp Fig. 7.

Results for different overlaminate

fillet stress, maximum overlaminate in-plane stress, and maximum overlaminate through-thickness stress, at a fixed load, for overlaminate thickness, gap, and backfill angle, respectively. Both

thicknesses.

centre- and two-clamp loading are considered, and each graph includes a series of curves for different fillet radii. Previous work on a similar problem14 had shown fillet radius to be the most

R. A. Shenoi, G. L. Hawkins

118

-----__________________

111

---__

t

10

15

Effect

18

of

25

20

Gap

30

35

40

45

50

35

40

+5

so

cap .1ze bn> on Deflection

.J

,:’ 16

,i ..i

..*’

14

70

Effect

CL xl

Effect

15

of

20

Gap

as M 33 cap *il. (mm) on Overlaminate

40

T-T

45

10

50

Effect

Stress

15

of

Is

of

25

20

Gap

5 Gap

30

ccp SIZC (m-m) on Overlaminate

In-plane

25 30 ~5 cap ain (mm) on Overlaminate

T-T

40

Stress

45

M

Stress

Two Clamp

Centre Clamp Fig. 8.

Results for different gap sizes.

significant geometric variable, and since it is the best representative of the scale of the joint, it is used as the ‘base’ for the results presented here. Firstly consider Fig. 7 and the results for the variation in overlaminate thickness. Backfill angle

and gap are constant at 45” and 20 mm, respectively, and fillet radius varies from 25 to 125 mm. Several features can be seen: 1. Overall deflections decrease with increasing overlaminate thickness and fillet radius. The

119

Petformance characteristics of a top-hat stifener

_______________ L

20

___-______________-_----_

zil_--_.. 0

5

1.9

K1

Backrll

Effect

of

Backfill

30

25

20

*Ias

Angle

3s

40

45

‘80

50

on

a

5

15

-.

20

29

B.dm

~~

Effect

Deflection

of

Backfill

,5

M

45-m

40

Pe-7 on DefleCtion

An!+

Angle

30

25

_____------------_______ ---__________----*

20

‘5 1---

g

-.---.“l..___

cz-zm

I

5

Effect

“0L ---i

-lo7

Effect

of

25

20

Bakrll

Backfill

An*

Angle

JO

35

40

(Dog on Overlominote

of

ncL ~-5 0

50

15

In--plane

Stress

_-I;

______ ._____._______ ________________________________1

Effect

Backfill

a

of

15

Backfill

Angle

20 Backfll

on

Stress

30 25 .4nrje (ce~

Angle

on

35

: fi in

40

Overlaminate

Fillet

45

50

h-plane

Stress

251

$6

___--. _________.____-_--____---2’ ‘r’ ______--__--_--i=’ A

_-L---_______,____,__._C

4

_,,

____---__________r______--~____-_

t Effect

of

Backfill

Angle

on

Overhminate

T-T

Stress

Effect

of

Bockfill

Angle

on

Overlaminate

T-T

Stress

Two Clamp

Centre Clamp Fig. 9.

Results for different backfill angles.

range of variation is small for centre-clamp and larger for two-clamp, as expected from Fig. 3, but this variation occurs in a relatively small region (the joint) of the top-hat

section and thus can have a significant effect on the stress distribution within this region. 2. The maximum principal stress in the fillet increases with increasing overlaminate

120

R. A. Shenoi, G. L. Hawkins

thickness for centre-clamp loading, but the range of variation is small. This is associated with an increase in the overlaminate inplane tensile stress (indicating a reduction in compressive stress) and an increase in the overlaminate through-thickness stress. Previous work on a similar problem16 indicated that failure in joints of this type is initially precipitated by delamination in the overlaminate (caused by high through-thickness stresses) followed by fillet failure. Failure never occurred in the plane of the overlaminate. In light of this it would seem that increasing overlaminate thickness, whilst being beneficial in reducing in-plane stresses, will reduce overall joint performance by increasing through-thickness and fillet stresses. 3. A similar trend is seen for two-clamp loading when the radius is small. However, for larger radii the fillet stress reduces as the overlaminate thickness increases. The range of variation is much larger. The in-plane stress in the overlaminate also reduces and the through-thickness stress increases, as with centre-clamp loading, and the same conclusion can be drawn. 4. For both centre- and two-clamp loading the fillet stress decreases with increasing radius, whilst the radius is small. Once the radius exceeds approximately 75-100 mm, however, the fillet stress level remains reasonably constant. This pattern is even more pronounced when considering the overlaminate through-thickness stress. Once the radius exceeds 100 mm stress levels increase, especially at higher overlaminate thicknesses. This indicates that excessively large radii can reduce the joint performance. Next consider Fig. 8 and the results for the variation in the gap between the stiffener web and the base plate. Overlaminate thickness and backfill angle are fixed at 2 layers of woven roving, and 45”, respectively. In a production context the gap size can be controlled within certain limits, but these will be variable depending on the complexity of the curvature of the base plate. The range of gap sizes considered in this study (lo-50 mm) was considered to sufficiently cover production variations. Significant results are: 1. Gap size has little effect on overall deflection for both centre- and two-clamp loading, indicating that overall stiffener performance

will not be affected by production variations. 2. For two-clamp loading the fillet stress reduces with increasing gap size until a minimum is reached, and as gap size increases further the fillet stress begins to rise. The optimum gap size is dependent on the fillet radius and generally reduces with increasing radius. For centre-clamp loading the same general trend is seen, except that as gap size increases to the higher values the fillet stress reduces again. 3. The overlaminate in-plane stress is little affected by gap size except at very small radii, where increasing gap size reduces the in-plane stress. 4. For centre-clamp loading the overlaminate through-thickness stress generally rises with increasing gap size to a maximum value, and then decreases again. For two-clamp loading this effect is less marked with little variation. Finally, consider Fig. 9 and the results for the variation in fillet backfill angle. The overlaminate thickness and gap size are constant at 2 layers of woven rovings and 20 mm, respectively. The backfill angle is considered as a variable here because in production it is a parameter that is very difficult to control, depending very much on the skill and consistency of the injector operator. Careful use with optimised nozzles should produce consistent 45” backfill angles with some attachment to the back of the web, but it was considered prudent to determine the effect of less than ideal production. Several salient points are shown: 1. Increasing backfill angle decreases the overall deflections slightly for the smaller radii with centre-clamp loading, but has no effect for any other geometry or loading condition considered. As with gap size this shows that production variance should not affect overall stiffener performance. 2. For centre-clamp loading the stress in the fillet decreases with increasing backfill angle. For two-clamp loading the fillet stress is unaffected by changes in the backfill angle except at high values when the fillet stress increases. stresses are 3. In-plane and through-thickness little affected by backfill angle except when the fillet radius is small. In these cases the stresses decrease with increasing backfill angle.

121

Performance characteristicsof a top-hat stiffener

5 CONCLUSIONS Generally the results of this work have shown that the new design of top-hat stiffener is a feasible alternative to the current practice. Moreover, the results indicate that a suitably refined joint should be able to at least match the overall performance of the current design without being prone to premature delaminations in the overlaminate. This, along with the potentially considerable savings in production costs, should ensure its further consideration. With the exceptions noted above, the ‘production’ variables (gap size and backfill angle) have a limited effect on the performance of the joint. Fillet radius and overlaminate thickness have significant effects on joint behaviour and can be optimised for a given application. Increasing fillet radius and reducing overlaminate thickness can overcome the problems of delamination in the overlaminate by reducing through-thickness stresses. This is in direct contrast to current practice. Finally it has been shown that the localised variations in the lay-up of the joint between the plate and the stiffener have very little bearing on the overall performance of a stiffened panel. This implies that the design of the joint can be optimised to improve its performance, without impairing the overall performance of the stiffener. The graphs included in this paper can be used for a design optimisation procedure providing that failure information exists for the materials used.

ACKNOWLEDGEMENTS The work outlined in this paper was funded by the SERC (administered through the MTD Ltd) and Vosper Thornycroft Ltd, Southampton, UK.

REFERENCES 1. Smith, C. S., Structural problems in the design of GRP ships. Proc. Symp. on GRP Ship Construction. The Royal Institution of Naval Architects, London, 1972, pp. 33-56. 2. Smith, C. S. & Dow, R. S., Interactive buckling effects in stiffened FRP panels. In Composite Structures 4, ed. I. H. Marshall. Elsevier Science Publishers, 1987, pp. 122-37. 3. Clarke, M. A., Stress analysis of a frame-skin connection in the MCMV. Internal Report, NCRE/L8/76, (Restricted), March 1976. 4. Dixon, R. H., Ramsey, B. W. & Usher, P. J., Design and build of the huIl of HMS Wilton. Proc. Symp. on GRP Ship Construction. The Royal Institution of Naval Architects, London, 1972, p. l-32. 5. Green, A. K. & Bowyer, W, H., The development of improved attachment methods for stiffening frames on large GRP panels. Composites, 12 (1981) 49-55. 6. Bashford, D. P., Green, A. K. & Bowyer, W. H., The development of improved frame to hull bonds for GRP ships - Phase 2. Fulmer Research Laboratories Ltd y;y;rt R871/1/July 1981, on MOD Contract NS7211/ 7. Bird, J. & Bashford, D., The use of flexible resin to improve bond connections in GRP ship connections. Proc. Composites 88, The World Conference on Composite Structures, Nice, 1988. 8. Gibbs and Cox, Marine Design Manual for GRP. McGraw Hill, New York, 1960. 9. Naval Engineering Standards, GRP ship design, NES 140, Issue 2, Undated. 10. Rules for Yachts and Small Craft. Lloyd’s Register of Shipping, London, 1983. 11. Rules for Building and Classing Reinforced Plastic Vessels.American Bureau of Shipping, New York, 1978. 12. TentativeRules for Classificationof High Speed and Light Craft. Det Norske Veritas, Hovik, 199 1. 13. Greene, E., Marine composites: Investigation of fibre reinforced plastics in marine structures. Report No. SSC 360, Ship Structure Committee, US Coast Guard, Washington, Contract No. DTCG23-89-C-20015, June 1990. 14. Shenoi, R. A. & Hawkins, G. L., Influence of material and geometry variations on the behaviour of bonded tee connections in FRP ships. Composites, 23 (1992) 335-45. 15. Users Manual for ANSYS, Swanson Analysis Systems, Houston, PA, 1990. 16. Hawkins, G. L., Holness, J. W., Dodkins, A. R. & Shenoi, R. A., The strength of bonded tee-joints in FRP ships. Plastics, Rubber, and Composites Processing and Applications, 19 (1993) 279-84.