LABORATORIUM VOOR
SCH EEPSCONSTRUCTI ES
TECHNISCHE HOGESCHOOL - DELFT
RAPPORT Nr.
BETREFFENDE:Low cycle
fatigue ofintersections of
bottom longitudinals of tankbarges. By ir. H.G. Scholte & ir. J.H. Vink.
r.
Sb i.p Structures Laboratory,
Deift University of Technology, Mekeiweg 2, Deift,
The Netherlands. Report
SSL 168
LOW CYCLE FAkFIGUE OF INTERSECÑONS OF BOTTOM LONGITUDIITALS OF TANKBARGES
by
ir. H.G. Scholte 1)
md
ir. J.H. Vink 2)
Deift Ship Structures Laboratory. Netherlands Ship Research Centre.
Contents. Summary Introduction Test specimen Test procedures Test results 5.1. Static tests 5.2. Fatigue tests 5.3. Fracture experiments
Conclusions and final observations
References Appendix I Appendix II
Figures nr. i - 50
See also addendum to rep.nr. 168
page i 2 3 4 6 6 8 17 19 22 A-II-1
-
1-LOW CYCLE FATIGUE OF INTERSEC1IONS OF
BOTTO1 LONGITUDINALS OF TANKBARGES by ir. H.G. Scholte and ir. J.H. Vink.
§ 1. Summary.
Lw cycle fatigue tests with full-scale structural sDec±mens have been carried out in the Ship Structures Laboratory at Deift on a 600 ton tension-compression-mchine. The tests were carried out on request of the Netherlands Ship Research Centre and of the n.y. shipyard and
engineeringworks De Biesbosch - Dordrecht.
The specimens represented the interconnection of longitudinal frames and transverse bulkheads constructed in normal shipbuilding steel. One specimen was provided with interrupted brackets another specimen
with through brackets. A third specimen contained a longitudinal
bulk-head. in addition to the longitudinals. All specimens had an overlap in
the bo±tomplate just below the transverse bulkhead. The overlapwelds
of overlapstrip to the bottomplate have been welded pat'tly with a
thick-ñess of6 mm and partly with a thickness of 5 mm.
Axial cyclic loading was.applied (P min./P max.
-i).
Prior to thefatigue tests some static tests were carried out. After fatigue testing the specimens were pulled to rupture
The way of testing as well as the number of specimens and the desired variation in structural details were determined in discussions between the n.y. shipyard De Biesbosch - Dordrecht, Bureau Ventas, the
Nether-lands Ship Research Centre and the Ship Structures Laboratory. A
con-tinuous help to the authors has been prof.ir. J.J.W. Nibbering (reader T.H.D.), who to a large extent also determined the experimental basi. Attention is given to the distribution of strain, the resistance to
fatigue and the effect of fatigue. ciacks upon the strength of the st'uc-ture. The report is ended with some conclusions and recommendations.
2
-§ 2. Introduction. .
During more than 10 years many pushharges have been built at n.y. shipyard
and engineeringiiorks De Biesbosch - Dordrecht.
Ali these barges had. conventional intersèctions of bottom longitudinals with flat bottomplates . However, as a result of reorganisation' and
improve-ment of production techniques, the yard likes to change the intersections in such a way that the bottomplates will be connected by means of an over-lapweld, instead of a butt joint. However, insufficient knowledge is avail-able about the influenceof such an overlap on the p1ace of the
inter-sections. Moreover, the barges,, to be built in the future according to
the new production-system, have to be accepted for classification for K1 and K3 cargo. So, it was thought necessary to gain more inforration about the strength of the intersections with overlap weld by means of experiments
on some full scale structural parts.
Two aspects had to be included in the investigation.
At first it was found important to get more detailed information in case of
the distribution of strain over the specimen; especially the points of
dis-continuity, the influence of the overlap upon bending effects in the bottom-plate and the E over the length of the intersection.
virtual
Secondly it was important tobe informed about'the lifetime of the structure
under low cycle fatigue loading. As the barges are used for inland navigation,
only the load amplitudes depending to loading and unloading of the barge will be of interest from a viewpoint of fatigue. With a lifetime of about 50 years, one voyage a week and 2 loadvariations per voyáge, it is clear that only
low-cycle high-stress fatigue is of interest. So it was agreed to restrict
the fatigue experiment to a maximum of 10.. 000 cycles per specimen and to
start the fatigue investigation with a stress amplitude for o . between
2 2 nominal
- kg/mm- and +8 kg/mm
In addition to an investigation concerning the overlap weld in the bottom-plate, also a comparison had to be made between intersections with trough-brackets and intersections with interrupted brdckots and finally the influ-ence of a longitudinal bulkhead had to be looked for. As moreover, for reasons of costs and time the number òf specimens had to be kept as low as possible, the investigation had to be restricted to one specimen of each
type; of course each specimen cöntains two testing parts.
In this way the experiments give not more than an indication about the fatigue resistance of the intersections and the result should be seen mainly as a
§ 3. Test specimen.
For the experiments three different specimens were available, all of them fabricated under normai working conditions by the n.y. shipyard and engin-eeringworks "De Biesbosch - Dordrecht". . In the first instance a.
test-specimen was preferred composed of bottomplate and transverse bulkhead
with 4 bottomlongitudinals , and in the middle a longitudinal bulkhead .
As this structure is very heavy and probably would cause many difficulties
in testing, it was decided to replace this structure by three other spec-imens..
First two specimens type A and B concerned the intersection of two bottom longitudinale (figure 3), without disturbing influences from other struc-turai longitudinal members. The difference between type A and B is found in the choice of the brackets. Type A has interrupted brackets, in
agree-ment with the most simple way of construction and fabrication and thus most preferable from a viewpointof economicalaspects for the yard. Type B,
as a through-bracket type, is preferred by the classification society. Where after strainmeasuremen-t in specimen A and B, very large strain was
found in the bottomplate at the end of the bottomlongitudinals,it was
de-cided after some discussion to modify specimen B with intent to decrease this large strain. Assuming that a circular cut out at the end of the frames would make the transition of the forces from the longitudinal into the
bot-tompiate less discontinuous, resulting Ip a decrease of the strain, two. frames were modified, as can be seen in figures 2 and 13.
Secondly a specimen type C (fig. 1i) was tested. This specimen was built up with two longitudinals and a longitudinal bulkhead in order to et
informa-tion whether a very rigid longitudinal strudtural part relieves the load
upon the rather flexible intersection of the bottorrilongitudinals, Based upon
the results of strain measurements with specimens A and B, it was decided to build up specimen (with bottom longitudinals) of the through-bracket
type.
When the tests should be restricted to the through-bracket type and the structure would fail in the bottomplate, there would he no information if
the interrupted bracket would have been. strong enough. With a limitation to the interrupted brackets and a failure of the bracket it would he uñknowi
if a through-bracket iou1d be (much) better. So the main reason for
includ-ing two types of brackets was to avoid a possible necessity for a second
experimental testing afterwards. -
-For similar reasons two types of overlap welds were used; one overlap with
a throat of 6,5 mm and one overlap with a throat of 5 mm, as shown in f
1g-ures 3 and 4. Classification rules require a throat thickness of 6,5 mm, being 70% of plate thickness. On the contrary, the yard prefers a throat thickness of 5 mm as in their welding practice has been proved that a weld
of 5 mm can be made in one layer without problems and independent of welding position. However, when the thickness of the weld is larger than 5 mm, it will be necessary to lay down the weld in three layers and this means that
the weld will be much more expensive; may b without necessity. A
restric-tion to only a weld of 6,5 mm throat thickness will not give any informarestric-tion if a weld of 5 mm will behave really worse than a weld of 6,5 mm. And if
there would be no difference in strength behaviour .there would be no need
for the weld of larger thickness. In relation to these considerations it
was decided to join the overlap strip with the bottompi.ates by means of a 5 mm weld at one side of the transverse bulkhêad and by means of a 6,5 mm
weld atthe other side.
For determining the properties of the material the yard delivered a strip
-150 X .12. The properties of this strip and also the properties of the
bot-tomplates are given in Appendix I. For determining the properties of the bottomplates, small testpieces were taken out of the edge of the bottomplate at halfway between transverse bulkhead and the end of the structure.
- Lt
--§ Li. Test procedures.
All specimens have been tested in the 600-tons tension-compression machine
of the Ship Structures Laboratory of the Technological University at Deift, (figure 2).
The investigation had to be split up into three parts; statical tests,
fatigue tests and loading to failure.
For all three specimens the experiments were started with some static tests to determine the distribution of strain over the structure. Three static tests were carried out. The first test concerned axial tensile- and com-pression load and enclosed two cycles with a maximum amplitude equal to the amplitude of the first series cycles of the fatigue load. For speciren A and B successively strain and E . were measured at loads of 0, -t-30,
virtual
-t-80, +130, -t-80, -t-30, 0, -15, -'-f0, -65, -L0, -15, 0 tonf. For specimen C the
steps were 0, -i-'-0, +130, -i-220, +130, -t-140, 0, --20, -65, -110, -65, -20, 0 tonf. (The locations of the strain gauges are given in figure 5, 6 and 7;
figure 8 shows the location of the points between which a measurement of E . is carried out).
virtual
The second and third static tests concerned bending, successively by means of a vertical load upon the top of the transverse bulkhead and by means of a horizontal load upon the topflange of the bulkhead. For all the specimens as well with the vertical as with the horizontal loading, measurements were
carried out at forces of 0, ±1500, +3000, -t-14500., + 6000, +3000, 0 kgf.
During these bending tests, the specimen could be considered as fixed in at one end, while the other end of the specimen was free to move in axial
direction.
As second part of the experiments the fatigue tests were carried out.
The fatigue loading represented more or less the loading of the bottomlongit-udinal of the barges due to the difference in longitbottomlongit-udinal still water bending moment for the empty and the fully loaded condition. The first part of the fatigue testing énciosed 3000 cycles with a stress amplitude for o .
. 9 -)
nomina-between a . -4 kg/mm and o +8 kg/mm .
-n mi-n n max
Besides in all fatigue tests the tensile component of the fatigue loading was taken two times as large as the compressive component (P mm/P max -b). The frequency of testing was between Li and 8 cycles per minute depending
on the magnitude of the applied load.
Returning to the agreements of the meeting on 16-11-1971 at Delft, Bureau Ventas proposed afterwards to lower the amplitude of the first 3000 cycles
to a . /a -2/+14 kg/mm2. However, the Ship Structures Laboratory
n mIn n max
preferred the original agreed loading, since it is very difficult to keep the upper- and lower value of the load amplitude constant with enough accur-acy just in that very low region of load capacity of the 600-tons test
machine. Besides, it was not expected that a nominal stress of -2 +14 kg/mm2 would cause any fatigue crack and in that case it would serve no useful pur-pose when 3000 cycles of -1 . ±8 kg/mm2 are being preceded by 3000 cycles
of half value. Moreover, when during the fatiguing the amplitudes háve to be
enlarged, there should be not too large differences between the
load-amplit-udes. So, finally, it was agreed to maintain the original program.
To get the nominal stress of -14/-t-8 kg/mm2 in the first test experiment for
specimen type A the first 3000 cycles were carried out with a load amplitude
of -65/-t-130 tonforce. After 3000 cycles had passed and no crack could be
observed, the load amplitude was increased to -80/+160 tonforces in order to
get a nominal stress of abt. -5/i-lo kg/mm2. The intention was to increase the load amplitude again when, a total number of 6000 cycles should have been
passed and still no crack initiation could be observed. However the first
crack was found after abt. 3750 cycles and after a total number of 14620 cycles the fatigue test was stopped.
For the specimen of type B the sane loading procedure was applied. Although after 2500 cycles a small crack was found, the amplitude was still increased after 3000 cycles, as otherwise a comparison between type A and B wòuld
become very difficult, if not impossible. Besides, a crack is, initiated and
already over a certain length proìagated before a crack will be observed, so that it would be not impossible that also for specimen type A the crack
initiation took place before. the first 3000 cycles had passed. Based upon
the preliminary results of specimen type A, the fatigue testing of specimen
type B could be continued to a total number of 5800 cycles with a crack
pro-pagation to a much larger extent than was the case with specimen type A. In consequence of a larger area of the transverse section, specimen type C was tested with an amplitude of -1101+220 tonforce for the first 3000 cycles and with an amplitude of
-1357+275
tonforce for the following cycles in order to have the same nominal stresses in the bottom section.The fatigue test of specimen type C was ended after a total number of 8500
cycles.
After ending the fatigue part of the test, all specimens were loaded under
axial compression up till the point of local buckling,, hOwever, without accepting a collapse of the specimen. For this purpose the total
lengthen-ing of the specimen was measured by means of a displacement transducer at
the end of the specimenat the point where theaxial load was applied.
As shown in figure 1 thi measured displacement was plotted on a X--Y-recorder
in relation to the applied load; the latter measured by means of pressure
dynamometers. Next to this the specimens were examined visually. In all these tests the compression load was increaed very carefully Until the dis-placement intended to grow even if the load was kept constant. At that moment the load was dropped down immediately. As the last step the specimens were
pulled under axial tensile load to failure.
As already mentioned above, strain measurements were carried out under sta-tic loading to determine the distribution of strain in the specimen,. Four types
of resistance strain gauges are used; all strain gauges of the fabricate
Tokyo Sokki. The rosettes are of the type PC-5, the torsional gauges, of 'the type PR-5 and the single gauges of t'ho type PL-5. So all strain gauges have
a filament length of 5 mm and a resistance of 120 1- 0,3 . Ónly the strain gauges nr. '41, 42, 51 and 52 on specimnen B are of the type PL-2, with a filament length of 2 mm and a resistance of 60 0.
For the. measurement of the E . four clip-gauge devices .weie used, as virtual'
developed in the Ship Structures Laboratory for C.O.D.-rneasurements. The
out-pUt of strain gauges, the clip-gauges and pressure dynamometers was measured by a digital measuring system, fabricate Peekel, composed of some 50-point connecting boxes, a power supply, a scanner and a digital voltmeter. The out-put was registrated by means of a high speed Teletype paper punch. The paper-tape was read out fo direct control, while later on the tape was used as input for an I.B.M. 60/360 computer to determine the Es-values, etc.
During the fatigue test 6 strain gauges of the ±eference section were con-nected to Peekel strain indicators and at intervals registrated with a
6-channel Rikadenki pen-recorder as an extra control in adjusting and cOntrolling
5.
Test results.
A.
Static tests.
The results of the. static tests are given in the figures 9 till 21.
For the axial loading the results are given in EC-values (kg/cm2) for a
tensile load of 100 tonforce. This could he done as all local strain,
measured with the strain gauges, proved to be practically propoitional
to the applied load. This inean that also in the heavily loaded zones,
even in the immediate vicinity of severe stress raisers, little or no
cyclic plastic straining occurred. This can be seen for example on the
computerplots for the strain gauges 13, 14, 15, 25 and 16 of specimen A
in the figures 22 till 26. The only exception in linearity was found fer
some strain gauges at the longitudinal bulkheadof specimen C, which was
due to the unflatness of thebulkhead.
Forthis reason the computerplots for the strain gauges 8V, SA, 9V, 9A,
23V, 23 A, 2t4V and 2A are included in this report in the figures 27 till
34.
When comparing the results of specimen A and B with the results of
specimen C it should he noted that all results are given in EE-values
for 100 ton axial tensile load. However, the area of a transverse
sec-tiori cf specimen C is much larger than the area of specimen A and B.
For a comparison based upon equal nominal stresses it will be necessary
to corrigate the values for specimen C with a factor of abt. 1,7. In the
same war, also the axial loads in testing applied upon specimen C were
chosen
1. 7 times as high as was applied upon specimen A and B
, 'in order
to get the same conditions undei' testing. In a first look, this factor
1,7 seems to be somewhat doubtful, as the nominal area of the transverse.
section of specimen C is 315 cm2
,that is abt .
twice the area of A aìd B
with a nominal area of 166 cm2
.However , there were two reasons for
taking a testload for specimen C, which was less thanabt. twice the
loading force for specimen A and B.
At first , just next to the transverse bulkhead the area are respectively
380 cm2 for specimen C and 225 cm2 for specimen A and B, thus a ratio of
abt. 1 7. Secondly and of much more importance, is the influence of the
bending aspect in the specimen. In fact, the variation of loads and
stresses in the bottom section of the pushharges are the results of the
differences in bending moment in the transverse section of the barge
during loading and unloading the barge. That
means that a tensile force
and stress in the bottom is accompanied at the
same time by a compressive
force and stress of the same order of magnitude in the
decic. So we have
not a constant value for strain and stress over the height of the
struc-ture. For the transverse section cf the barge we find the neutral axis at
a height of abt. 22'40 mm above the bottomplate. The neutral axis of
spe-cimen A and B lies 143,6 mm above the bottomplate, while the neutral axis
of specimen C lies about 283 mm above the bottomplate. So the
mean value
of stress in specimen C has to be (22LI0_283)/(221o_4I4)
0,89 times the
mean value of specimen A and B, assuming that also the bending effect is
simulated in testing of specirnenC. This results in
an applied load for
specimen C of (315/166) x 0,89
1,685 times the applied load for
speci-men A and B. With the tapering-off of the longitudinal bulkhead towards
the ends of specimen C the distribution of stress and
train over the
nominal transverse section is brought into better agreement with reality,
resulting in a nominal stress of lower value i
specimen C as is the case
for specimen A and B, while stress and strai.n in bottomplate
and
longitu-dinals of specimen C are the senne as for specimen A and B.
From figures 9, 12, 13, and 18 it can be seen, that the ineasured strains
trans-verse bulkhead and the end of the longitudinal, have much higher values than anywhere else in the structùre. The high strains are the results of axial strain and load bending strain and it can already he concluded but
of the static results. that this point will he thb weakest of the struc-turc; that is to say that in this region the first crack i'nïtitation can be expected.
Comparing smecimen A with specimen B, there seems to be no large differen-ce. In the brckets of type A more bending occurred, hut strain values are not growing really high. However, the strain measured by gauge np. 12
on the bracket of specimen A is about 1 times the strain at the sanie place of specimen -B. This is due to local bending in the bracket and from
this point of view bracket type of specimen B may be preferred. Looking
for the bottoinpIate at the strain gauges 15/25 of specimen A and the
values of strain gauges at the same locations of specimen B, we see that the maximum strain value is about 10% higher for specimen A. So at a
first look snecimen 13 seems to be a little bit better than specimen A, hut this should be handled very carefully.
Concerning the modification of specimen 13, where after te first static measurements two longitudinals were modified hy cutting out a circular hole at the end of the webplate of the longitudinal, it pointed out that the modification was not an improvement of the structure. The mean value, the bending value and the maximum value of strains measured at- tIe end of the longitudinals were enlarged with about 30%.
For determining the E . the displacement between two points as
indi--virtual
--cated in figure 8, was measured in relation to the applied load.
- In accordance with the strain measurements also the displacements proved to be proportional to the applied load, so that the measured elongation
per load of 100 tonforce also are no-ted in the figures 9, 12, 13 and 18. The E . , based upon the mean value of the elongation at points C
virtual .1
C2, C3 and C1, was found to be 1,52 x l0 kg/mm2 for specimen A.
For specimen B a value of E . was found to be 1,60 x l0 kg/mm2
be-- virtual -
-fore the modification. Due to the modification of the longitudinals, the
E . decreased to a value of 1,50 kg/mm2. virtual
For specimen C also the same elongation measurement has been carried out. However, it is not easy to transfer this measured elongation into a
value for E . , as it is not weil done to define what in this case has virtual
to be understood by E . . However, it can he seen as very pleasant,
- virtual
-that it is proved to be possible to calculate the elongation very
accura-tely out of the measured strain values of the strain gauges. This is
important, since it is therefore possible to calculate the E . or,
- virtual
what is of more importance, the deformation orelonation over any
-length and at any place you like.
The measured elongations are all compared with calculations-, based upon-the distnjbijtiìon of strain in the specimen as measured with the strain gauges. The measured- - and calculated values were in entire agreement with a
maxi-mum deviation of abt. 3%. Some examples of this calculation as weJ.l as
some controls upon the applied axial load are given in appendix II. As already mentioned in chapter
1,
the - specimens werè also applied to avertical load and to a horizontal load upon the - top of the transverse
-In figure 2 it can he seen how the loads were ap1ied, Just as was the case with the axial lading, the measured strain values were all exactly
pro-portional to the applied, load. The results are pldtted in figures 10, 1'4,
15 and 19 for the vertical loading, while the results for the horizontal loading are given in figures 11, 16. 17 and 20. For both cases the results
are given as E-values (kg/cm2) per 5000 kgf. For comperason of these results for specimen A and B with the results of specimen C again it
should be noted that the structure of specimen C is such heavier s the
structure of specimen A and B in this respect.
B. Fatigue tests
The results of fatigue tests are given in the, figures 35 till '4'4.
As mentioned already in '4 the first 3000 cycles of fatigue testing were carried out with an applied axial cyclic load amplitude of -651+130 tf for
specimen A and B and a load amplitude of -1101+220 tf for specimen C, cor-responding with a nominal stress amplitude in the bottomplate of -'4'-f-8 kg/mm2. When the first 3000 cycles in testing specimen A had passed and no crack could be observed, it was decided to increase the applied load with 25% and testing was continUed with a load amplitude of -80/-i-160 tf,
according to' the agreeded test program. In order. -to have the same test
conditions also the load amplitude of specimen B was increased up till -801+160 tf, after 3000 cycles had passed, while the load amplitude of
specimen C was enlarged up till -135/+275 tf.
At a total number of 3750 cycles for specimen A first cracks were observed,
haying some length already. Fatigue was stopped at a total number of '4620
cycles for specimen A, wen no one of the cracks had developed over a depth
equal to the plateth'ic1ness.
in specimen B, first cracks were found at 2500 cycles. Unless this
presen-ce of cracks the load amplitude was changed at 3000 cycles in order to be able to compare the results with those of specimen A.
Fatiguing of specimen B was continued Until cracks were developed
of-con-siderable greater extent than in specimen A, in order to get better
insight into crack growth characteristics. At a total number of 5800 cycles fatiguing of soecimen B had beer, stopped when crack A 3, fig. 35 and 38,
had developed over some length a'depth equal to plate thickness.
In testing specimen C again first cracks were found before 3000 cycles
of --1101+220 tf had passed. However, according to the test procedure of specimen A and B. the load amplitude was changed. into -135/+275 tf for the
next 3000 cycles. At 6000 cycles total it was decided to continue
fatigu-ing as the cracks were s-till rather small; may be to small to he sure of
having fracture at a tension load below 600 tf. The lòad amplitude was now maintained upon -135/+275 tf. Fatiguing was stopped at' 8500 cycles total, lust before collapse during fatigue could b.e expected due to crack
B.'4 (fig. 36 and 39).
Figures 35 and 36 show growth of crack length ' in relation to number of cycles. The cracks are split up into some types and provided with a code in relation to their location in the structure (see fig. 35-39).
The code encloses a letter A or B and a figure 1 up to 6 and is specified
'as given below:
A : fatigue crack initiated at the toe of the
overlapweld at the top side of the bottomplate, aocI developing
-downwards into the bottomplate
B : fatigue crack initiated in the weld-undercut of the (overhead layed) overlapweld at the underside of the of the overlap stiip and developing upwards, into the
-9-i-2 : fatiguecrack is initiated near the end of a longitudirial at that side of the transverse bulkhead, where 6,5 mm overiapweldsare used
3-14 fatiguecrack is initiated near the end of a longitudinal at that hide of the transverse bulkhead, where 5 mm overiapweldsare used
b etc. : fatiguecrack along one of the above mentioned weidments, but outside regious as indicated by thefigures 1, 2, 3 and 14.
If necessary, the code may be preceded by a letter A, B or C indicatirg
the type specimen.
Cracks of type Ai-14 and Bi-14 were discovered at almost the same number of cycles, but they behaved in a quite different manner.
Cracks of type Al-14 gradually propagated in length at both cracktips as
-well as in depth as the number of cycles increased, while cracks of type
Bi-'4 initiated over a larger extent of length, but propagated less in
depth.
Cracks of type A5 etc. are of less importance and became visible in a
later stage of fatigue.
The explanation of these fenomena is quite logical. Due to the interruption
of the longitudinals the largest stress and strain in the bottornplate
will occur between the end of the longitudinals and the transverse bulkhead,
with a stress raiser at the toe of the longitudinal and a tapering off
towards the transverse bulkhead (see appendix II figure AI:[). So, looking along the entire length of the overlapwelds the largest hart of' plate
stress will be found just at the end of the longitudinals in thé thinner bottomplate, see figUre 21. The effect of bending stresses dueto the
overlap will also depend upon the- location in length and breadth with
respect to the end of the longitudinals, as well as to the thickness of the plate. Unfortunately the direction of bending comes through in such a
way that stress and strain in the overlapstrip and bottomplate at the toe of the overlapwelds are increased considerably. In line with the interrupted
longitudinals the bending stress in the thinner bottomplate will be higher than in the bulkhead strip, (see figures 9, 12, 13, 18 and 21) as the length between bulkhead and overlap divided by thickness of strip
is larger than the length between end of longitudinal and overlap divided by thickness of bottomplate. Away from the line of longitudinais the length from bulkhead to overlap does not change while the length from overlap to endpoint can be thought to increase very sharp. This results in a sudden decrease of bending stresses at the toe of the weld at the side of the longitudinal. However the bending stresses at the toe of the weld near the transverse bulkhead will hardly change over the length of the weldment
(which is also found in figure 21).
-Due to the above mentioned effects the stresses will be maximum in the
bottornplate in the line of longitudinals. In that particular region crack
initiation and largest propagation can be expected As is clear these effects will exist for cracks of code A, and indeed the results for- these
lo
-For the overlap welds at the underside of the bottomplate close to the transverse bulkheads , and with a larger distance to the ends of the
longitudinals , where cracks with COd B are found, the situation is more
complicated.
The hart of plate stress is relatively small because of increased
thickness of overlap-strip with respect to bottomplate; mean levels are related with factor 0.9/1.2.
The hart of plate stress, when looked along the weidment, does not show
a very high peak in way of the longitudinals because of the larger distance of : this overlapweld to the end of longitudinals.
These effects can very clear be seen in the figures of appendix II, where E values for hart ¿f plate are given for the bottomplate along the line of longitudinals, and in figure 21 for distribution along the weidment.
The bending stress alöng this weldment will show some raise when
further away tram the line-of longitudinals while the value of this stress in line with the longitudinal is less than in the hottoniplating.
The resulting stress distribution along the toe of this weidment f
the strip will accordingly show much less a peak in way of the
longitudinal-intersection and the mean value is much less than the peak value at the toe of the overlapweld of the bottornplate just in way of the longitudinal.
So, crack initiation and propagation couldbe expected to be of much
lower value in this region. However, the conditions at the toe of the
weld are much worse as a result of the weld-undercut due to ovehead
welding, which effect is also present along the entire length of the overlap-weld. For this reason, initiation of cracks with code B can also be
expected in an early stage. Initiation will be more extended along thd
weld ìcnt as the difference in conditions in way of longitud inals with respect
to the rest of the weld is little. Propagation in depth will be ess for
this cracks with code B than for cracks with code A. This is due to
the fact that after initiation of cracks the cracktip conditions, are no longer influenced by the weld-undercut as stress raiser.
At the other side of the overlapweld, in the bottomplate on the line of longitudinals the field of the stess raiser due to the interruption of the longitudinal has considerable longitudna1 and transverse dimensions with respect to crack dimensions and stress raiser due to the presence of a crack in this particular region has to be superposed upon the
high stresses that already exist. This gives a situation in which cracks
with code A will propagate faster than cracks wit« code B. 0f course, for cracks with code B the first initiation point along the weldrnent
can be expected in the neighbourhood at the longitudinals. The propagation
into the depth will he fastest in this region too as can be expected but slower than for cracks with code A.
Except for crack B.-i of specimen C the results are in fui.ly agreement with the above given explanation.
For speciment A sorne measurements of length of cracks type Al_11 were made,
but the number was not sufficient to get insight into crack growth
characteristics as fatigue was stopped at an early stage.
For speciment B and C more crack length measurements were carried out and
the results are plotted in figures 35 and 36.
These figures show that crack initiation has taken place before 3000 cycles.
For specimen A there is only one line of crack growth; the other cracks are given with end-points. For specimen B and C the curves of cracks
After changement of load amplitude when
3000
cyLes. have been appliedcrack groth rate starts to increase strongly for most cracks type A partly due to the increased load amplitude. This high crack growth tate however is of limited duration and a period of 1essr drack growth
starts. This is the moment when the crack length reachés beyond the
region of peak stresses due to longitudinal interruption. Sàon a
minimum, crack growth rate is reached and then a continuous increase
of crack growth rate follòws due to a not furthermore neglectible
reduction of cross rectional area.
It seems tht crack growth of speciment B is somewhat faster than for specimen A and C but one has to remember that cracks Al and A3 of specimen B, which show high crack growth., are located near longitudinal-ends of
modificated type where increased stresses exist withrespect to
longitudinal-ends without moditication. Cracks A2 and A-t of specimen B,
however, show crack growth of the same order as specimen A and C.
Comparison of mean crack lengths of different specimen in relation with measured Ec-meanvalues in the region of longitudinal-ends is mad in
table I.
Table I Comparison of mean crack lergths
i) Ec mean values for a load of 100 tonf for specimen A and B and
170
tf for specimen C.2.) Mean value at upperside of bottomplate..
3) Hart of plate valúe.
Lt) Cracks at longitudinal ends with modification.
Cracks at the endoflongitudinals without transverse brackets. Cracks at the end of longitudinals with transverse brackets.
Length of A3 at Lt600 cycles is guessed at 57 mm with the aid of figure
35.
These values are something extreme because of crack
A2 of
specimen C. From this table it follóws ' that crack growth is in good relation to Ec values at the upperside of bottornplate near the end of the longitudinal.The modifications of specimen B are disadvantageous with respect to maximum stress level as well as crack propagation. Some of the final cracks type A in specimen B and C extended over all the thickness of the plate for some length. The numbers of cycles at which was observed that the bottom of the crack reached the other side of the plate, are given in figures 35 and
36,
indicated by NTH.Specimen Crack
no.
mean crack length at
'4600 cycles 5800 cycles Statical I 2) E 1.) ii
3)
AAl... .4
79 min -4820
1882 BAl,A3 6)
124 mm 178 mm 5916 1805 A2,A4 75 mm 7) 109 mm 4617 1829 CAl,A4 4)
75 mm 94 mm4850
1947A2,A3 5)
42 mm 8) ', 68 mm.8)
4660
181712
-Figures 37, 38 and 39 show final crack dimensions for all cracks tyne A,
while particularities of these cracks are shown in the pictures of figires -1-0, '1-1 and 42.
Figure '40a shows crack' Al of specimen A at the side of the 6,5 mm weld,
looking from bulkhead to longitudinal. This is a normal fullr developed
crack at '4620 cycles. Figure '4Db shows cr'arlc M of specimen B at the
side of the 5mm weld, looking from.bulkhead to longitudinal. This is
an other normal fully developed crack at, 5800 cycles. The dark line in the
bottom of the crack to the right is a shearlip due to break out the
piece afterwards for examination.
Figure '4la shows crack AL1 of specimen C at the side of the 5mm weld,
looking from bulkhead to longitudinal. This shows a normal crack which is propagated over the full plate thickness for some length at 8500 cycles. See also cracks in weld on toe of longitudinal which ki,nd of cracks also occured in way of Al and A3 of specimen C in something lesser extent,
but nowhere in speimen A and B due to less cycles, even not in M of
specimen B (figure '40b) which shows much weld-undercut in the longitudinal. Figure '41b shows crack Al of specimen B at the side of the 6,5 mm weld, looking, from bulkhead to longitudinal. This figure shows a lamination in the crack bottom. Number ot cycles N 5800.
Figure '+2 shows some not fuÏly developed cracks of type A:
Figure 42a shows crack A2 of specimen A at the 6,5 mm weidside, looking
from bulkhead to longitudinal at '4620 cycles.
Figure '42b shows rack A3 of specimen A 'at the 5 mm weidside looking from longitudinal to bulkhead at '4620 cycles.
The horizontal cracks in this piece are caused when the other part was
broken out of the plate for examination of the crack surface.
Pieces of figures '-1-2a and 42b are photographed with inclination of the
cracks. BotJ pieces are somewhat damaged because of the great forces for
breaking out them.
The lamination found in specimen B, crack Al, may be seen as a defect
which is rather normal for materials of properties as used. The fatigue
crack is somewhat ce1ayed for growing into the depth by this lamination. The length of this crack is not excessive as crack A3 ot specimen B, which is comparable, has more length. For these short term experiments, a
lamination has not so much influence even when at, present in the regions
of longitudinal ends, in practice, when a corrosive liquid canpenetrate
into the lamination, it becomes more dangerous,
To get some indication about the difference in behaviour of cracks type A3,'4 at the 5tnm weld side with respect to cracks typeAl,2 at the 6,5mm
weld side.see table II for final crack lengths, crack areas and' measured Ec values.
- 13
-Table
IITotal final lengths and areas on éither side. of
bulkhead arid measured Ec values.
Notes 1), 2) and 3): see table I on page 11.
In most cases cracks A3,Llat the 5 mm weidside have little more length and some more area than cracks Al,2 at the 6,5 mm weidside due to some
more depth at the 5 mm weidside. Looking at strains of specimen B at the 6,5 mm weidside these seem to be something larger but cracks are worse.
at the 5 mm weidside. However, in the
5 mm weld itself, strain is in
excess of strain in the 6,5 mm weld (straingauges 51 and 41).
It is hard to indicate the difference as
siignifidant, butit looks like
that there is little tendency to somewhat better behavior of
6,5 mm welds
with respect to. cracks of type Al-4.
Though sorne cracks of type B are plotted in figure 36 they are considered in another way as cracks of type Al-4, r
Figure 43 is a picture of an example ot 'cracktype B at. N 4620 for
specimen
A at 6,5 mm weldside. It i's clearly shown in this picture thatthis kind of crack is not only concentrated near the longitudinal-bulkhead-intersection, which place is marked by straingauge 28, These cracks
developed over a much greater extent and are initiated in the bottom
of the weld-undercut due to overhead welding which is found in the
bulkheadstrip. Tese cracks, having little crack opening due to little penetration and being hidden in the bottom of the weld-undercut,, are most difficult to determine in an early stage during
fatigue tests. Length of these cracks during tests can hardly be mentioned
with good accuracy. They appeared on
both sided of the bulkhead in the
bulkheadstrip.
Besides cracks of type Al-4,at the ends 'of longitudinals, hardly any crack of 'type A5etè. appeared because of lag of weld-undercut due to
downhand welding. Only crack AS specimen B (figure 38), and in specimen C near the longitudinal bulkhead some cracks are found.
TabLé UI gives 'some information about 'the distribution of these cracks,
because itgives total final lengths in percentages of weld lengths and final number of cycles for each specimen as found after a thoroughly
inspection of the specimen when experimentswere finished.
Total final length
and area of cracks
Ali-A2 at 6,5 mm
weldside
Mean E
values 1)
Total final length
and area of cracks
A14-t-A3 at 5 min
weld-side
Mean Evalues 1)
I 2) II 3)
I 2)
II 3)
Specimen A
NTot 462086-F73159 mm
2 358-i-l2348l mm --80-t-77l57 mm
275228303 mm
. 4820 188» Specimen B NTot 5800 l60f78238 mm 37-1'l50524 mm 5526 1860 l0+195z335 mm ' 49i+'825J3l6 mm46 154
L' Specimen C NTot 8500 2l9+l55374 mm 2 9l5i-44ll356 mm -250+206456 mm 2 l276±8l42090 mm 4755 18801.) Mostly near crack tips of cracks A1-'4.
Crack A5 figure 38 and vicinity. Near longitudinal bulkhead mainly.
LI.) These values may be extra high as much length
of these cracks is found in the vicinity where final fracture cracks stopped while material over there
shows much elongation and cracks were opened. Outside cracks s drawn in figures 37, 38 and 39. First value is mean value, second value is max. depth in way of B1,3 and the third in way of B2,i..
One can see now that these cracks appear to be somewhat deeper at the
5mm weldside, unless initiation' is hardly the same for both sides.
From figure 36 it clearly can be seen that crack BLI. of specimen C, starting
very slowly, has become unstable at about a number of 8400 cycles total, which is shown in table IV to coincide with the moment that the bottom
of this crack reached the other side of the bulkheadstrip. Table IV Unstable growth of crack B'4 of specimen C
1) This is exclusive a length of 500mm where a crack of little depth is
found afterwards.
Outerside of hull deep weld undercut
Innerside of hull, no weld-undercut
5
Outside regions of longitudinals
Bli-B2
6,5mm weld
B3+B4
5mm weld 6,5mm weld ' 5mm weld
% i depth % 1 depth % i depth % i depth
Specimen A N '4620 98 % 0,38mm 1,40mm 1,00mm 92 % 0,67mm 1,00mm 1,20mm 1) i-10% -i) +10% -Specimen B N 5800 83 % 0,53mm -1,0 mm 73 '% 0,91mm 0,50mm 2,50mm i- 5% -2) i-20% 4) max.l,5 mm Specinien C N 8500 68 % 2,20mm 1,50mm 10,5mm 79 % '4,8 mm -12 mm i-3,5% -3) +22%
-Cycles Length at underside of strip Length at upperside of strip
7000
85mm
'0
'7830 140mm 0 7975 163mm 0 8350 300 mm 0 8470 not measured ' 125 mm 8540 500 mm 1) 320 mmTable III Length in % of weldlength where cracks are found and depth of cracks B and A5etc, initiating from weld-undercut.
-. 15
-It has to be remembered however that the moment whn crack B- of specimen C became desastrous, due to unstable crackgrowth atN 8!400, is well beyond
the moment of acceptable cracks of type A, as cracks Al, A3 and A'4 of
specimen C had developed over the full plate thickness before N = 7830
(see figure 36).
Final fatigu.e crack dimensions of crack ß14 of specimen C is shown in figure 39. At the same moment crack B2 of specimen C had developed in this
specimen to dimensions as given in figure 39 too. This once more gives an indication that there is only little difference with respect to weld
thickness. It can further be concluded from this that, unless initiation of these cracks is found hardly over all the weld length, the propagation
into the depth is faster in the regions near longitudinal intersections,
which is clear. This is found in something lesser extent in specimen A and B too.
Regarding foregoing results of cracks of type B it is hard to entitle these differences as significant hut it seems that 6,5mm weld behaves somewhat
better due to less stress concentration at the toe.
During fracture of specimen A it became clear that in the welds, connecting the interrupted brackets to the bulkhead, fatigue cracks existed which were not determined during fatigue experiments as they had not yet reached
t.he weld surface.
At a tensile load of 250 tons during fracture test these cracks were opened
over the lower part of the brackets.
Looking at figure 9 it follows that in the interrupted brackets, which are
overlapped to the longitudinals, bending occurs, resulting in higher
stresses in that side of the bracket where the longitudinal is overlapped.
Now looking at the final fractures of the brackets, cracktypes according
to figure 4'-i- are found.. At the side of overlap, with higher stresses, type
Cl crack is found over all the height of bracket, with a mean distribution of crack kinds over the throatthickness according to specification a; C 1a
At the other side of the brackets a mixture of two fully different crack
types is found: cracktype C 1, with specification b, is mainly located in the upper part of the bracket and cracktype C 2 mainly in the lower part;
see detail of cracks C in bracket i and bracket 2 figure Distribution
of crack types C 2 and C ib over this side in percents of bracket height is given in table V.
Table V
Crack types C i a and 'C 1 b, having somewhat different mean distributions of crack kinds over throat-thickness, are of formas given in the
specification of figure :
- A first brittle initiation part over K% of throat-thickness.t.,
- A second fatigue part over L% of throat-thickness.t.
- At last a fibre or mixed fracture over M% of throat-thickness t. Mean values of K, L and M for cia and Clb,cracks are according.
Bracket 1 . ' Bracket 2 type C2 .
65mm
20% height '151mm
47% height type db 80% . 53%to table VI
Table VI
It is concluded that a brittle initiation crack has developed from the throat over some 16% of t already in an early stage of fatigue or
before. 'This crack can appear over there due to following combined factors; Ti) Inclusions in the weld material.
Bad material properties due to quick cooling of weld material. Stress concentration due to quick cooling of weld material.
L)
Strain concentration when loaded.All these factors besides the fourth one are of lesser magnitude when
further away from the corner and.the brittle initiation crack may stop. During fatigue the crack propagates over some LI0% of throat thickness. Finally a rest fracture of fibre or mixed type over some 1f5% of throat thickness took place during fracture test.
It can be expected that most fatigue length will be at present near the lower part of the bracket at the side where the longitudinal overlaps,
but this is not confirmed when. looking at the surfaces of the fractures.
Interesting is that several brittle initiation parts are found in the lower parts of the welds at the other side of the bulkhead too.
Crack type C 2 is as drawn in figure 4: In this case the ciack starts
in the 'H.,A..Z near the throat of the weld, runs over sorne length through the weld-material, crosses through the H.A.Z. and reaches the plate material.
As here the properties are good it is turned back over some 90 degrees
to the plate surface at the toe of the. weld through the 1I.A.Z. again. At first it was thought o,f a fracture due to lammellair tearing but
microscopic inspection pOinted out that it was not, the case.. Both brackets
fractured in the same way and crack parts of type C 2 must be caused by
the same effect which only can be found in the crack sequence of the. welds connecting the brackets to the bulkhead.
Looking, at the effect of the longitudinal bulkhead it can be reviewed that
first and worse fatigue cracks appeared near longitudinal intersections in specimen C. At the end of the fatigue experiments hardly any fatigue crack was found near the longitudinal bulkhead (only some 12 mm at N 8350 at the 5. mm weld-side on each side of the longitudinal bulkhead along the toe
of the inner overlap-weld). From the point of view of fatigue the bulkhead
construction is safer than the longitudinals.
The transverse brackets on one side of the longitudinal bulkhead may have.
effected the results of that longitudinal which they are connected with. Indeed at the side where no brackets are placed crack growth is faster, but it has to be regarded that in specimen A and B a somewhat faster crack growth is found in the same side of the specimen too; see table VII.
01a
cib
K 16 16
L '+3 . 38
17
-Table VII Final crack dimensions at either sid of specimen.
Comparing fatigue resistance of longitudinals of specimen C without
transverse brackets (Crack Al,4) with the unmodificated longitudinals of specimen B (cracks A2,4) shows little difference as is shown already in
table I on page 11. So the influence of the longitudinal, bulkhead upon fatigue behavior of an adjacent longitudinal intersection seems to be
negiectible
when no transverse brackets are connected with thelongitudinal.
C. Fracture experiments
Fracture experiments were carried out after fatigue testing of the. specimen.
First the specimen were compressed until a maximum compressive load wìs
reached at which negative eongation takes place without a further increase of compressive load. This could directly be observed in a recording in which P (tf) was plotted against the displacement of the traverse of the 600
tons tension-cômpression machine. As an example to illustrate these kinds
of recordings figure '45 shows the character of one of them.
Having reached the point of unstability the compressive load was released
and followed by tensile loading up till fracture.
Table VIII gives Pmin and P W.T.S. for the three specimen:
Table VIII
Length
and areaof Al
+ AQLength and area
of A2
1- A3relative propertie
1. and A1
'r and Ar
-1./'r
Al/Ar
86 -t- 77 163 mm 73 80 153 mm
1,07
Specimen A 358 + 75 '433 mm2 123 228 351 mm2 160-t-.l'40300mm78+l95273mm
1,10 Specimen B 374 + '491865 mm2 150'
+ 825 975 mm2 0,94 219 + 250 469 mm 155 + 206 361 mm 1,30 Specimen C 915 +1276 ;2l91 mm2 '4'4l + 814l255mm
. 1,32Pmin Puit.tensile foregoing total number of cycles
A - 175 t 395 t '4620 cycles
B - 190 t 340 t 5800 cycles
18
-Figure '6 gives the final fractures. From this figure it becomes clear that cracks of type B, having not caused limitation of fatigue life with respect to acceptable crack growth in the hottcm, in all casés were involved with final fracture in the bottomplate. Only in specimen A final fracture was not restricted to the bottomplate, but accompanied
and initiated by fracture of the brackets.
Only specimen A showed a total fracture while for specimen B and C the crack in the bottomplate was stopped. In specimen B it stopped because of extra elongation when fatigue crack A,3 opened. In specimen C it may
have been the influence of the longitudinal bulkhead to stop the crack. For specimen B and C the brackets did not fracture.
At 250tf tensile load brackets of specimen A showed opening of cracks
that already existed.
In specimen B some little cracks were found afterwards in the weld at the
end of the longitudinal where the bracket is connected.
It is important that, unless cracks of typ A caused limitation of number of fatigue cycles because of crack depth, no one of these cracks
failed during fracturetests and did not influence ultimate strength. Cracks of type B did so more. In specimen C crack B,14 had become already
unstable during fatigue testing and initiated final fracture.
During the fracture tests the loads were applied stepwise and for each load the strains were measured by the digital measuring system. As a result of computer handling with this information plots of epsilon with respect to load are available for the straingaeges. From these nsight can be got
into fracture sequence and behavior of the structure during the period
of high loads as well as behavior under compressive load. Sonic of the characteristic plots are collected in a separate addendum to this report.
-- 19
6. Conclusions and final observations.
In all specimen fatigue cracks. are initiated at the toe of the
overlapuelds in the bottom after about 2000 cycles of a
loadamplitude, corresponding with a nominal stress amplitude of -41+8kg/mm2 for the bottomplate in the undj'turbed section. The initiation and propagation of the fatiguecracks at the toe of
the overlapwelds at the upper- or inner side of the bottomplate is the result of the high strainvalues between transverse bulkhead and the end of the longitudinals, partly due to the stress- and strainraising effect of the interruption of the longitudinals and partly due to the large bending effect of the overlap.
The initiation of fatigue cracks at the toe of the overlapwelds at the under- or outerside of the overlapstrip is especially the result of the stress- arid strain-raising effect of the large
weldundercut. in combination with the bending effect of the overlap, while the stress and strain-raising effect of the inter-.
ruption of the longitudinals is of less' importance.
The geometry of the weidments and the applied welding techniques are
very important. In this respect the attention is drawn upon the
weld-undercut, due to overhead welding, which weldundercut can be seen as primarily responsible for the initiation of the cracks mentioned in point 3.
5 For all specimen the local strainvalues, determined in staticàl
measurements before and during fatiguetests, proved to be pràcticall.y
proportional to the applied load. This means that in tse heavily loaded specimen, even in the immediate vicinity of severe
stres-raisers little or no plastic training occurred.
Interrupted brackets show a larger horizontal bending than throug
brackets, resulting in higher local strain and stress. Partly due to this bending effect, but primarily due to the kind of welding
connection between brackets and transverse bulkhead, fatigue cracks
were initiated in this weld. Although fatigue resistance of the
structure was not restricted by these fatigue cracks, the collapse .of the structure under loading to rupture was initiated in the fatigue
cracks in the welds between bracket and bulkhead.
Comparing 6,5 and 5 mm wej.dments of overlapwe.ld with r'espect to their
fatigue resistance, the 6,5 mm weld showed a little hit better behaviour, notwithstanding the larger strainvalues in bottomplate near the 6 mm weldments (see strainmeasurement of specimen 13 figures 12 and 13). However, the difference could not be pointed out to be very significant.
It should be noted that no remarks can be made with respect to the
influence of the bulkhead stiffeners, as the difference in stifines of the intersection on both sides of the transverse bulkhead may be related to the behaviour of the overiapwelds.
20
-The modification of specimen B proved to have an unfavoura.ble influence
upon the resulting strainvalues and fatigue-resistance.
The total loadvariations applied during the fatigue tests within a
couple of days, will be spread over about 50 years in reality. So
partly due to the short time of testing
but primarily in consequence
of the absence of corrosive elements, the influence of corrosion was
not included in the investigation. Fio:ever, it should be noted that
corrosion fatigue
rack will develop easily as the fatigue resistance
cf unprotected structures in a corrosive environment is very low.
/s could be expected the intersection of the longitudinal bulkhead
behaved much better as the intersections of interrupted longitudinals.
With
respect to the influence of the longitudinal bulkhead upon the
adjacent bottom longitudinals, the results show a tendency that the
longitudinals adjacent to the longitudinal bulkhead and connected
to
the bulkhead with transverse brackets tere in. little advantage compared
with the other longitudinals.
Howeve; the differences are very small and by far not significant.
The measured values of the E virtual over the intersection of the
longitudinals showed a reduct.ion of about 30% with respect to E normal
for the specimen with interrupted brackets and about 25% for the specimen
with through brackets. However, it must be realised, that these
figures are arbitraly as the E virtual depends strongly
upon the distance
over which the value is determined. From calculations based upon the
measurements of strain distribution over the structures, it was found
that E virtual can be determined with good accuracy.
Looking at the results of fatigue testing äs presented in figures 35
and
36 it seems that the fatigue resistance is low. However, the results
are
presented for a loadamplitude during the first 3000 cycles corresponding
with a nominal value of _L/±8 kg/mm2 in the bottorripart of the
structures
at the underturbed sectional area, while the loadamplitude
was enlarged
up till -51+10 kg/mm2 for the following cycles. Assuming that the real
loadarnplitude as it will be present in practice,
may be of lower value,
the results are translated for other lower values of loadamplitudes. This
translation is based upon earlier results from experimental work carried
out in the shipstruc-tures laboratory, which investigation is published in
report 82S. (7)
In this report crackgrowth results are given fbr initiation of cracks
2(crack area is zero), crack areas of 100 mm2 and crack
areas of 500 mm
as
a function of nominal. stresses (strain x youngs modulu) and number of
cycles. Some of these results concerned fatiguecracks in bottomplates
between transverse bulkhead and ends of longitudinals
at the place of
intersections. Although these cracks developed at the toe of the
weld of
longitudinals to bottomplate, the. conditions
are the same and a good
comparison is thought to be possible. The results as presented in figures
6 and 8 of report 82 S are used for the translation of fatiguecrack results
of intersections of the pushbarges. The donfiguration of the specimen
as
shown in figure 6 of report 82 S is most according to the specimen
of
the pushbarges and therefore this figure
s used primarily. However, as
the results of the specimen showed in figure 8 of report 82 S
are somewhat
unfavourable with respect to an extra-polat ion into lower stress- and
21
-For translation of the results it is assumed that the lines for certain
cracklengtb have the
sarre character as the lines for the. crack area.
The first part of the translation enclosed areplacement of the results
of figures 35 and 36 by a new diagram with a line fo mean uppervalues of
fatiguecrack
growth. (figure7 )
In the same figure also
the
growth of fatiguecracks is given for a constant
ioadarnpììtude of -5/-t-10 kg/mm2. The translations were made by
means of two
diagrams based upon the results from figures C and 8 of
report 82 S
(see figures '-18
and
L9 ).
From this diagram the numher of cycles could be
determined for different cracklc-ngth as function of loadap1i.tude.
For predicting the crackgrowth for a ioadarnplitude of O/+'4 it is assumed
that the influence of the compressive parts of the loadamplitude
uponfatigue damage is half as much as for the tensile part of the amplitude.
In this way a prediction is given for loadamplitudes of -5/-t-10,
-41+8,
-2/L
and O/' kg/mm2 .
The results are given in figure 50
.Comparing the different specimen and the testresults
,it seems evident
to analyse once more this way of construction J)efore it will be used in
normal production. Although the differences between the interrupted- and
the throughbracket
structures are not very large, it must be concluded
that the throughbracket is prferab1e to the interrupted bracket.
Perhaps also the welded connection between :i.nterrupted bracket and
transverse bdlkhead can he improved, using a larger throatthickness of the
weld or using a K-type weld. However, a larger thickness of the weld
oxa K-type weld will be more expensive from a viewpoint of cost and time.
When using an interrupted bracket, welding must be carried out very
carefully.
With respect to the thickness of the overlapwelds a
6,5
mm. weld seems to
be preferable.
However, the differences
in
resistance against fatigue between the 5,5 and5 mm weld arc very small. Although the bottomplate fractured
in the
5 mmwelds under loading, to rupture,
it
is still the question if the 6,5 mm weld is much stronger than the 5 mm weld. Concerning the initiation offatiguecracks in the overlapwe.lds
cannot
he said that the 5 mmweld is
worse. Of more importance is the effect of local bending due to the
overlap and the strainraising effect due to we].dundercut. Therefore it is
important to check if these effects can be
reduced
or eliminated. Withrespect to the local bending, it can be expected that the large strains
due to the bending effect of the overlap
can be reduced largely
when the
thickness of the overlapstrip is increased
up till about 20 mme
Due to the larger stiffness of the thicker ovenlapstrip
the angular
desplacement at the overlap is reduced, resulting in smaller
bending strain
between overlap and end of longitudinal. Bending strain in the overlap is
decreased even more, due to the larger thickness. Also the mean value of
strain in the overlap is decreased in
consequence of the larger thickness.
This is also very important in connection with the
weidundercut dUe
ooverhead welding. In order to decrease the strain- and stress-raising
effect of the weldundercut, perhaps the weld
can be improved by means cf
a TIG melt, grinding or peening. From a viewpoint of cost deep grinding
will not
be attractive, but comparing approximate costs peenirig seems tobe much less expensive. (2)
Another improvement of the overlap might be
to enlarge the overlap in
such an extent, that the overlapwelds are located under the longitudinal
22
-Finally attention has to he paid to corrosion prevention. Concerning the
requirements that the barge must be classified for K1 and K3 cargo, it seems that a corrosive environment can he expected, Looking at the pollution
of the inland waterways we can assume that the possibility of. a Corrosive
environment wilinot be restricted to the innerside of the barges. At one side it may be advisable to look for improvemen-ttechniques of weldments
a. peening of TIG welding and at the other side corrosion pretection must
be considered. A normal corrosion preventing paint, however,, is. not
sufficient as it can only help to prevent initiating of corrosion fatigue
cracks. Sut the protection against corrosion fatigue is finished when cracks have started already due to other (strainraising) effects. In fact
this corrosion protection is of really importance in the regions of
longitudinal intersections and overlapweids. For this places, where a large chance exist that fatiguecracks are initiated, a corrosion protecting
medium should he used, with so large an elasticity that this surface coating will be kept intact, even if a small fatiguecrack develops uìder the
coating. In this way a kind of rubbercompound e.g. at a thiokoll base, may be of good help. 0f course such a compound must have a good resistance
against the corrosive cargo. References
Publication of the Netherlands Ship Research Centre Report 82 S, April1966.
!ILow_cycle fatigue of steel structures"
(,an experimental investigation with full scale ship structurai components)
Fatigue of Welded Structures Conference, 6-9 july, 1970.
J.D. Harrison, F. Watkins and mrs. P.H. Bodger.
The fatigue strength of welded joint in high strength
i