Documente Academic
Documente Profesional
Documente Cultură
Abstract: An important feature of ship design is the ability to describe the structural behaviour of the hull and to
accurately predict its ultimate strength. Research on the ultimate strength of hulls has been carried out by many
experimental research groups. Recently, the consensus reached is to simulate the behaviour of the hull by loading a
box girder up to its ultimate strength. A box girder was tested at the Royal Military College of Canada. The objective
of this experiment is to study the structural behaviour and compare the experimental results with the predictions of two
computer codes. The construction of the model follows typical hull construction methods. The girder was subjected to
pure bending until failure occurred. Provisions were taken so that collapse would occur due to buckling and not to
plastic failure. Residual stresses and initial geometrical imperfections were measured and considered in the analysis. In
previous publications, details of the design, fabrication, and loading were presented. In this paper, the experimental
results are described and discussed.
Key words: box girder, ships hull, bending, ultimate strength, residual strength, residual stresses, initial imperfections.
Résumé : Un élément important dans la conception de bateaux est l’aptitude à décrire le comportement structural de la
coque et de prédire avec précision sa résistance ultime. Des recherches sur la résistance ultime des coques de bateau
ont été effectuées par de nombreux groupes de recherche. Le consensus atteint récemment est de simuler le
comportement de la coque en chargeant une poutre caisson jusqu’à sa résistance ultime. Une poutre caisson a été testée
au Collège Militaire Royal du Canada. L’objectif de cette expérience est d’étudier le comportement structural et de
comparer les résultats expérimentaux avec les prédictions de deux programmes informatiques. La construction du
modèle est conforme aux méthodes de construction typique d’une coque. La coque a été soumise à une flexion pure
jusqu’à ce que la rupture survienne. Des dispositions ont été prises de sorte que l’effondrement serait dû au flambage
et non à la rupture plastique. Les contraintes résiduelles de même que les imperfections géométriques initiales ont été
mesurées et considérées dans l’analyse. Dans des publications précédantes, les détails de la conception, de la
fabrication et du chargement ont été présentés. Dans cet article, les résultats expérimentaux sont décrits puis discutés.
Mots clés : poutre caisson, coque de bateau, flexion, résistance ultime, résistance résiduelle, contraintes résiduelles,
imperfections initiales.
Fig. 2. Cross-sectional configurations of a ship’s hull (a) and the box girder counterpart (b).
b σy
[4] β =
t E
Dimensions
where Et is defined by
The box girder was 10 m long, with a cross section of
Et 362
. β2 960 mm by 720 mm (Figs. 3–5). The thickness of the plating
[5] = ; 0 ≤ β ≤ 19
.兾 p
. + β兾4
was 4.55 mm in the test section and 9.53 mm in the remain-
E 131 ing part of the box. In order to represent as closely as possi-
ble the Canadian Forces frigate, a ratio of approximately 1:2
Et
[6] =1 ; β > 19
.兾 p is used for the following four parameters: thickness of the
E plating, cross section of the longitudinal stiffener, spacing
An iterative process is required to obtain the failure stress between the longitudinal stiffeners, and spacing between the
of the panel, as the failure stress is dependent on the reduced transverse frames. The details of the cross section for the
effective plate width, and the effective plate width is defined box girder, its longitudinal stiffeners, and the transverse
by the failure stress. frames are shown in Figs. 3 and 4.
Dow and Smith (1986) used a two-dimensional (2D) non-
linear elastoplastic finite element analysis to establish the Materials
strain–stress curves for the stiffened panel. Initial distortions Four different grades of mild steel were used. These are
of the plate and stiffeners and the residual stresses were ac- the types of steel commonly used in warship construction.
counted for in their program FABSTRAN. Table 1 presents the results of standard tensile tests; each av-
erage is based upon the results of three to five samples.
These results show, as expected, that the yield stresses are
up to 40% higher than the nominal value and the modulus of
The formula of Faulkner (1975) and the computer pro- elasticity is approximately 200 GPa. All other details are
gram FABSTRAN were used to calculate the ultimate published elsewhere (Akhras et al. 1995a; Shyu et al. 1995,
strength for the box girder. Using the formula of Faulkner 1996).
Fig. 6. A panel with the longitudinal stiffeners and the transverse frames.
Fig. 8. A global view of the box girder and the loading frames. Fig. 9. Micro-strains (in kN·m) at failure of 14 outside strain
gauges.
-1119 842 -197 -851
Top
VNW2 VSW1
North South
-288 Mid-span of the west segment -555
VNW1 VSW2
* 745
HBW1 HBW2 HBW3
Bottom
PM1
PM3
W PW2 PMA PMB PMC PMD PM4 PME PMF PMG PMH PE2 E
PM5
PM7
cause, after the first failure, the box girder was flipped over Fig. 11. Initial residual stresses for the cross section where
and tested again. The 2220 kN·m bending moment, which is buckling occurs.
the maximum value that could be produced by the two point
loads of 750 kN each, is designed to exceed the theoretical
fully plastic flexural strength of the girder by 37%.
Loading frames
The design of the experimental setup is detailed in Akhras
et al. (1995c). The construction followed very closely the
plan outlined by Akhras et al., and Fig. 8 shows a photo-
graph of the setup with the box girder. The outriggers and
the loading beams were used to ensure that the damaged box
girder was tested using the same loading pattern.
20 48 76
11 13 27 39 41 55 67 69 83
91 32 60 88
3 10 14 26 29 38 42 54 57 66 70 82 85
9 15 25 37 43 53 65 71 81
W 2 8 16 24 30 36 44 52 58 64 72 80 86 E
7 17 23 35 45 51 63 73 79
4 90 89
1 6 18 22 31 34 46 50 59 62 74 78 87
40 68
12 28 56 84
5 19 21 33 47 49 61 75 77
A B C D E F G H I J K L M
9 21 30 42 51 63
4 14 16 25 35 37 46 56 58
69 68 67
2 5 13 17 23 26 34 38 44 47 55 59 65
W 6 12 18 27 33 39 48 54 60 E
3 24 45 66
1 7 11 19 22 28 32 40 43 49 53 61 64
15 36 57
8 10 20 29 31 41 50 52 62
Fig. 13. Initial geometric imperfections for the cross section To calculate the residual stresses, the strain measurements
where buckling occurs. were multiplied by Young’s modulus of the longitudinal
-0.5 -0.2 1.1 0 1.8 1.8 4.0 stiffeners, which has an average value of 206 GPa (Rawson
and Tupper 1983; Shyu et al. 1996). The maximum calcu-
lated surface residual stress due to fabrication is a compres-
Top sion of 21 MPa. As an example, Fig. 11 shows the residual
HTW3 HTW2 HTW1
-1.2 -1.6 weld stresses for the middle cross section of the west seg-
ment where buckling occurs.
-1.9 VNW2 VSW1 -1.5
North South
0 Note: All positive values indicate elevations 0
toward the inside of the box girder Initial geometric imperfections
-0.3 VNW1
Values are in mm VSW2 -0.4 Initial geometric imperfections are induced by welding
and fabrication. The magnitude and direction of the initial
-0.7 0.5
HBW1 HBW2 HBW3 geometric imperfections of the box girder can significantly
Bottom affect the theoretical analysis and computer solution. In fact,
they can explain the difference between one type of failure
0.47 0.6 1.5 0 -0.6 -2.5 - 0.7
and another in numerical modelling. Consequently, the box
girder was constructed with considerable attention. Initial
geometric imperfections were measured using an automatic
surements Group Vishay, were used to resist this heat and level with a parallel-plate micrometer to enhance the reading
special installation procedures were needed. After welding precision. The locations of the measurements are numbered
the longitudinals, a second set of 40 standard Showa strain as in Fig. 12 and were taken for the four sides of the model.
gauges were installed, 30 in the middle of the reverse side of A total of 256 locations were measured (53 on each side
the webs of the stiffeners, and 10 on the flanges in the mid- panel and 75 in the top and bottom panels). The largest ver-
dle of the test section. tical imperfection in a cross section was 4.7 mm of node 84
Readings were taken at three phases of the box fabrica- in the bottom panel which is within the limits permitted in
tion: when the longitudinal stiffeners were cut with no weld- the specifications. As an example, Fig. 13 shows the varia-
ing, when the stiffeners were welded to the plating, and tions of the initial imperfections for the middle cross section
upon completion of box fabrication. of the west segment where buckling occurs.
Fig. 14. Time–displacement curves for three potentiometers located on the failed cross section.
Fig. 16. Moment–displacement curves for potentiometers on the top of the box girder.
Fig. 17. Moment–strain curves for a strain gauge on the top of the box girder.
Fig. 18. Moment–strain curves for a strain gauge on the bottom of the box girder.
Ultimate strength
The purposes of preliminary tests are to check the global Because all the preliminary tests were successful and the
performance of the complete experimental setup (actuators, behaviour of the box at the beginning of the loading process
supports, loading and distribution beams, loading frames, was linear, a loading rate of 31 kN/min was chosen for the
etc.), test the data-acquisition systems recording the reading beginning of the loading process and a much slower rate of
of the gauges, and verify the linear behaviour of the box 6 kN/min for the remainder of the experiment.
girder before loading it to failure. Three preliminary tests To provide more readings of the data collection systems,
were performed. In the first, the loads in the actuators were three pauses of 2 min each were allowed when the loading
increased until the bending moment reached 310 kN·m reached a bending moment of 931 kN·m for stage one, 1066
(which corresponds to a total load of 210 kN); the bending kN·m for stage two, and 1199 kN·m for stage three. Thereaf-
moment reached 621 kN·m (420 kN) for the second test and ter, the loading continued until failure occurred when the
932 kN·m (630 kN) for the third test. The results of the three cross section of the west segment of the test section buckled
preliminary tests were very satisfactory. They showed that and the loading reached a value of 836 kN for a bending
linear load–displacement and stress–strain behaviours were moment of 1238 kN·m. The central plating defined by the
in accordance with engineering calculations. longitudinal stiffeners and (or) by the borders of the box
Fig. 19. Moment–strain curves for three gauges on the side of the box girder.
buckled first. The loads released from the buckling of the girder. Figure 14 shows the time–displacement curves of the
plating were redirected to the longitudinal stiffeners, which three potentiometers PW1, PW2, and PW3. As shown in
finally buckled laterally, leading to the bending of the box Fig. 10, these potentiometers were located on the deck of the
western part of the box girder where failure was first ob- the potentiometers. As in Fig. 14, the same remarks hold for
served. The three curves in Fig. 14, which are almost super- the time–displacement curves of the seven potentiometers in
imposed, are divided into three segments. The slope of the Fig. 15. These potentiometers were located on the middle
first segment represents the loading rate of 31 kN/min used part of the deck (PM1–PM7 in Fig. 10).
in the beginning of the loading process, and the steeper Figure 16 gives the moment–displacement curves for five
slope of the second segment the loading rate of 6 kN/min potentiometers located on the deck of the box girder: the lo-
used in the last part of the experiment. The superimposed cations of PW1, PW2, and PW3 are as described earlier;
segments of the curves indicate that the displacements of the PM4 is located in the centre of the middle part of the deck;
three potentiometers were practically the same and thus no and PE2 is located in the centre of the east part of the deck
twisting, distortion, warping, or local buckling had devel- (Fig. 10). Figure 17 presents the moment–strain curves for a
oped during the loading process until buckling occurred. strain gauge installed on the top of the box girder in the
This observation corroborates the assumption that a plane compression zone, and Fig. 18 presents the results from a
cross section of the box girder stayed plane during loading strain gauge in the tension zone. Finally, Fig. 19 displays the
until failure. The third segment is flat and indicates failure: moment–strain curves for three gauges installed on the side
the three final readings represent the final displacements of of the box, one in the middle of the box at the neutral axis
location (OSW2), one at mid-distance of the upper part of (1) The ratio of the ultimate moment predicted by the
the box (OSW1), and one at mid-distance of the bottom part computer code FABSTRAN to the experimental moment
of the box (OSW3). Until the buckling load had been was 1.05, which suggests that, in general, overestimation of
reached, OSW2 registered no strain, for all intents and pur- the hull’s strength will result if this computer code is used. If
poses, as expected for a neutral axis, whereas OSW3 and the second computer code, which does not take in account
OSW1 behaved linearly in tension and compression, the initial imperfections and residual stresses, is used, then
respectively. These results clearly show that the box girder the ratio is only 1.01.
behaved similarly to a beam and collapsed due to pure bend- (2) The experimental residual strength of the damaged
ing which generated, in the compression zone, the buckling box girder in the reversed flexure direction was also deter-
stress required to fail the plating and the longitudinal stiffen- mined and was found to be 73% of the experimental ulti-
ers. The photographs in Fig. 20 show the buckling areas mate moment.
viewed from various angles and, as expected, several un- (3) In the first test, practically no twisting, distortion,
symmetrical semi-half-waves. warping, or local buckling had developed in any cross sec-
Again, buckling occurred as anticipated and the results are tion of the box girder during the loading process until buck-
very close to those generated by the computer models. The ling occurred. This corroborates the assumption used in the
theoretical ultimate moment Mu, without taking into account analysis of a box girder subjected to pure bending and stat-
the effects of the initial residual stresses and imperfections, ing that a plane cross section will stay plane during loading
was 1250 kN·m; FABSTRAN, the computer code which until failure (Fig. 19).
takes into account the initial residual stresses and geometric (4) The buckling mode of the plating panels had, as ex-
imperfections, predicted an Mu of 1300 kN·m, and the exper- pected, the shape of unsymmetrical waves delimited by the
imental result was 1238 kN·m. longitudinal stiffeners (Fig. 22).
(5) The concept of residual strength should be investi-
Residual strength gated further to improve the knowledge of damage assess-
Another objective of the project was to evaluate the resid- ment and the general behaviour of the structure.
ual strength of the damaged box girder. After failure, the
box was flipped over and tested again. As stated previously,
the outriggers and loading beams were designed and used so
that the damaged box girder could be tested using the same Financial assistance was kindly provided by the Depart-
loading pattern. ment of National Defence of Canada. The authors would
The loading and data-acquisition methods were the same like to acknowledge and thank the RMC Structural Labora-
as those used in the first test. Again, to guarantee more read- tory staff, the RMC Machine Shop staff, and CFB Kingston
ings of the data-collection systems, three pauses of 2 min Base Maintenance staff for their excellent work on the pro-
each were allowed when loading reached a bending moment ject.
of 228 kN·m for stage one, 452 kN·m for stage two, 674
kN·m for stage three, and 896 kN·m for stage four. As the
load was steadily increased, the buckled plating and longitu-
dinal stiffeners, which were now in the tension zone, Akhras, G., Shyu, C.T., and Ellis, J.S. 1995a. A box girder to
straightened up slowly before the compression zone failed in evaluate the ultimate strength of a ship’s hull. Proceedings, An-
buckling. The experimental maximum moment Mu reached nual Technical Session, Structural Stability Research Council,
was 913 kN·m, which is 74% of the experimental ultimate Kansas City, Mo., pp. 137–146.
moment of the structure. Photographs of the flipped box are Akhras, G., Shyu, C.T., and Ellis, J.S. 1995b. Ultimate strength test
shown in Figs. 21–23. model of a ship’s hull. Proceedings, Annual Conference, Cana-
dian Society for Civil Engineering, Vol. IV, Ottawa, Ont.,
pp. 403–410.
Akhras, G., Shyu, C.T., and Ellis, J.S. 1995c. Ultimate strength of
ship’ hull: I. design. Royal Military College of Canada,
A box girder, which simulates the behaviour of a ship’s Kingston, Ont., Civil Engineering Report CE95-2.
Caldwell, J.B. 1965. Ultimate longitudinal strength. Transactions,
hull, was designed, built, and tested at the Royal Military
Royal Institution of Naval Architects, 107.
College of Canada. The objective of this experiment was to
D’Arcangelo, A.M. 1969. Ship design and construction. The Soci-
study the structural behaviour and compare the experimental ety of Naval Architects and Marine Engineers, New York.
results with the solutions of two computer codes. The test Department of National Defence. 1981. Welding specification for
section of the girder was subjected to pure bending until HMC ships, D-49-003-003/SF-001. Department of National De-
failure occurred. The scantlings of the section were about fence, Ottawa, Ont.
half those of a typical Canadian warship, but the overall Department of National Defence. 1993. Structural practices stan-
cross section was about 1/15 that of a full-size frigate. The dard for CF surface ships, D-03-002-008/SG-003. Department
model was designed, as required, to prevent any other mode of National Defence, Ottawa, Ont.
of failure from occurring before local buckling of the plating Dow, R.S. 1990. Testing and analysis of a 1/3 scale welded steel
and stiffeners. Residual stresses and initial geometric imper- frigate model. Admiralty Research Establishment, British
fections were measured and considered in the analysis. The Ministry of Defence, Dunfermline, Fife, Scotland.
results clearly indicate that the test was a success and its ob- Dow, R.S., and Smith, C.S. 1986. FABSTRAN: a computer pro-
jectives were achieved. gram for frame and beam static and transient response analysis
(nonlinear). Admiralty Research Establishment, British Ministry Smith, C.S., and Dow, R.S. 1986. Ultimate strength of a ship’s hull
of Defence, Dunfermline, Fife, Scotland, ARE Report TR under biaxial bending. Admiralty Research Establishment,
86205. British Ministry of Defence, Dunfermline, Fife, Scotland, ARE
Faulkner, D. 1975. Compression strength of welded grillages. In Report TR 86204.
Ship structural design concepts. Edited by J.H. Evans. Cornell
Maritime Press Inc., Cambridge, Md., pp. 633–712.
Hughes, O.F. 1983. Ship structural design. John Wiley and Sons,
New York. a: length of the stiffener
Mansour, A., Yang, J.M., and Thayamballi, A. 1990. An experi- b: stiffener spacing
mental investigation of ship hull ultimate strength. Transactions, E: Young’s modulus
Society of Naval Architects and Marine Engineers, 98: 411–437. Et: tangent modulus
Ostapenko, A. 1981. Strength of ship hull girders under moment, Mu: theoretical ultimate moment
shear and torque. Proceedings, Ship Structure Committee, p: ratio of the stress at the beginning of the inelastic region
Society of Naval Architects and Marine Engineers Extreme
to the yield stress
Loads Response Symposium, Arlington, Va., 19–20 October
rc: radius of gyration of the combined plate and stiffener
1981, pp. 149–166.
Rawson, K.J., and Tupper, E.C. 1983. Basic ship theory. Longman,
R: residual stress reduction factor
London and New York. t: plate thickness
Shyu, C.T., Ellis, J.S., and Akhras, G. 1995. Ultimate strength of β: plate slenderness ratio
ship’ hull: II. Fabrication sequence. Royal Military College of λ: column slenderness for the panel
Canada, Kingston, Ont., Civil Engineering Report CE95-3. σf: inelastic failure stress
Shyu, C.T., Wheaton, G.G., Gillespie, D.S., Ellis, J.S., and Akhras, σy: material yield stress
G. 1996. Ultimate strength of ship’ hull: III. Tensile tests. Royal η: factor of the welding type and shakedown effects (η =
Military College of Canada, Kingston, Ont., Civil Engineering 3.75 for most naval ships)
Report CE96-1.