Fabrication and Fatigue
Failure in Aluminum
Close attention to design and best practices
can simplify construction and extend
the working life of a high-speed aluminum boat.
by John Kecsmar
Graphics courtesy of the author
(except where noted)
Above—Good welding practices coupled
with sound design of structural elements
go a long way in minimizing the fatigue
failure that’s common in many aluminum
boats, especially those that see hard
service in rough conditions.
56 Professional BoatBuilder
C reating the complex structure and
graceful curves of an aluminum
boat requires the profound transfor-
mation of flat, square, virgin plate. It
is rolled, bent, forged, bashed, scored,
heated, welded into shape, and
painted. Done incorrectly, any one of
these processes can seriously compro-
mise the fatigue life of the resulting
structure. Such failings can be exacer-
bated by original designs that over-
look the simple details needed to
accommodate the properties and
peculiarities of the material.
The relationship between aluminum
fabrication practices and fatigue failure
is complex. While I cannot possibly
cover all its aspects in a brief article, I’ll
highlight some of the common causes
of failure in aluminum boats that are
too often overlooked during planning
and construction.
To better understand why such mis-
takes are made and how to correct
them, let’s take a look at a common
fatigue failure in a series of high-speed
aluminum powerboats I worked on in
the early 1990s: cracking in the alumi-
num structure in transoms and around
waterjets.
At the time, the popularity of water-
jet propulsion was booming, and with
it came a troubling rise in these
cracks. They were common enough
to almost be expected. The problems
have now been well documented and,
 Figure 1. FEA Model of Transom and Waterjet
in most cases, remedied. Builders and
operators discovered that corrective
action early on could prevent the
need for more costly repairs, which in
turn prompted more research into
these fatigue-related failures.
For the teams designing, building,
and repairing these boats, the imme-
diate questions were: Why were the
boats cracking? And while it was clear
that the structural cracks were related
to vibration of the waterjets, engine
girders were subjected to similar vibra-
tions: Why weren’t they failing, too?
Answers weren’t self-evident, so
before specifying effective repairs that
wouldn’t repeat the problems in the
original structure, engineers were
tasked with performing a detailed
analysis of the mechanics of the failures
by applying very basic finite element
analysis (FEA). Then they strain-
gauged a waterjet transom and structure
during sea trails to validate the FEA
model.
Today, understanding a structure
subjected to a range of stresses is
Inspection of the damaged boats
revealed classic signs of fatigue failure
around the load paths identified in
the FEA.
Fatigue
In this context, fatigue is best
described by metal fatigue expert
L.P. Pook: “a failure of a metal under
repeated or otherwise varying load
which never reaches a level sufficient
to cause failure in a single applica-
tion.” In simple terms: When you
apply a force to a structural member,
it will bend and become stressed. So
long as the stress from static loading
is below the material’s yield stress, designers and fabricators to recog-
the member will not fail. If you now
repeat this loadingFigure many
3-1.times,
Fatigue Limits can
although each specific load applied properties, and avoid the sort of
does not overstress the member, the
simple repetition can initiate a failure
owing to the dynamic
or cyclic nature of the Figure 2. Fatigue Limits of Steel and Aluminum
loading.
In troubleshooting
30
our transom failures,
Semi-range of Stress x 10 MPa
In contrast, aluminum has no
fatigue limit. Constant exposure of
aluminum to repeated loading will
ultimately lead to failure, no matter
how low the actual stress magni-
tude experienced for all practical
purposes.
Faced with numerous failures, we
were beginning to think that alumi-
num was a poor choice for a light-
weight boatbuilding material; it
seemed cursed. It turns out that
many of our problems—and those of
other builders—were caused by
applying steel design and building
methods to aluminum, a very different
metal. Steel, a forgiving material, has
been around for centuries, and the
techniques and procedures for work-
ing it are highly developed and
widely employed. As a boatbuilding
material, aluminum is just decades
old, so correct, consistent aluminum-
fabrication methods are not as com-
monly shared. It is imperative that we
teach best practices for aluminum to
fabricators, especially those who are
accustomed to building in steel.
Basic Aluminum Fabrication
During fabrication, steel is not
affected and/or changed to the same
degree as aluminum, but aluminum
has the undeniable advantages of
light weight and resistance to cor-
rosion. So, how do we reconcile
these differences? It’s important for
nize aluminum’s characteristics so
they of
maximize
Steel and Aluminum its desirable
failures that plagued our waterjet
structures and transoms.

readily possible through even more or any fatigue failure

sophisticated FEA, which is often of an aluminum boat, 27
standard at the design phase (see there’s an important dis-
“FEA,” Professional BoatBuilder No. tinction to be made 24
78, and “Fine-tuning with FEA,” PBB between fatigue in steel Mild steel normalized
No. 133). Figure 1 shows a typical and fatigue in alumi-
FEA model of a transom and waterjet num. Below a particu- 21
structure. This analysis indicated loca- lar level of stress, steel Al Cu alloy
tions of higher stress, though not nec- reaches its fatigue 18 2014A T6
Endurance strength
essarily unacceptably high. Areas limit. No damage or at 50 million cycles
identified on the FEA were consistent loss of strength will
15
with locations of the actual failures, occur below that
but the stresses were all below the fatigue limit, regard- 105 106 107 108 109
designed allowable yield stress of less of the number of Endurance (Cycles)
the structure. So why did they fail? cycles (Fig. 2).

June/July 2012 57
Quality Control
Material-specific quality assurance (QA) is the key to
consistently successful design and construction of
aluminum boats. Whether you’re the designer, the plater,
aluminum, and helps you avoid the common errors
caused by not differentiating between the two materials.
You may ask, “Is strict quality control really that neces-
or the welder, your awareness of the alloy’s physical sary? I’m only building a small yacht; won’t it add time
properties and limitations is the first step in recognizing and cost? Surely this is only for large commercial builds.”
what can compromise an aluminum structure, what can Regardless of size or complexity, once the boat enters
Figure 4. Rotary-Saw
go wrong Cutting
in build, and Temperatures
ultimately how you in canChip and
avoid Plate
service, simple unforeseen vibrations from numerous
such faults. That understanding also clarifies the sources will expose any flaws unintentionally built into
necessity of different fabrication processes for steel and the hull due to lack of attention to detail during design

Let’s start with a few fabrication

Figure 3. Rotary-Saw Cutting Temperatures in Chip and Plate practices that best accommodate
aluminum’s principal characteristics.
Avoid marking the plate with scrib-
1,256 ing tools; these leave marks on the
(680) surfaces and can create a slight flaw
tc = 0.036" (0.91mm) where a crack can develop and
1,292 (700)
1,310 (710) become the site of crack initiation
1,274 1,328 (720) under high loads. Similarly, don’t
(690) 1,328
write on aluminum stock with pencil,
as the carbon in it is higher up (more
1,238 noble) on the galvanic corrosion
(670) series.
1,346 (730) If aluminum is too coarsely cut, its
rough surface can result in potential
1,364 (740)
Temperatures sites of crack initiation. Cutting with a
°F (°C) 1,382 (750) rotary saw can create considerable
1, heat at the blade tips (Fig. 3). This
1, 22
0
1, 202 (6 heat buildup can locally reduce the
18 ( 60
X 1, 6 strength of heat-treatable 6000-series
16 4 (6 50 ) Heat
6
(6 40)
) aluminum as well as annealing a
30 strain-hardened 5000-series non-heat-
1,148 (620) )
Figure 5. Effects of Notching or Scoring treatable alloy.
Y Another desirable property of alu-
Z minum that differentiates it from steel
is that it requires no special tools for
t = 0.024" (0.61mm)
bending. It is important to ensure that
the bending tools are free of irregular-
ities, which could cause marks or
Figure 4. Effects of Notching or Scoring score the surface of the plate. The
effects that notching or scoring have
on aluminum are shown in Fig. 4.
45 Lab air—smooth (as-received) Ranging from minor to major, notches
Lab air Lab air—Kt = 3 edge notched 300 in the plate can reduce the fatigue
Maximum Stress (ksi)

40 Lab air—Kt = 12 edge notched strength by as much as 75% from its

35 Salt water—smooth pristine condition.
6061-T4 sheet specimens, R = 0.1
30 Kt = 3 Clearly, the way you handle,
200
machine, form, bend, and strike the
MPa

25 Salt water
Kt = 12 alloy will affect its fatigue strength.
20 Small steel particles from fabrication
15 100 tools—such as a steel hammer or steel
rollers—could unintentionally be
10
embedded into the plate, where they
5 can act as a site for crack propagation
0 0 or suffer from corrosion leading to
103 104 105 106 107 108 109 stress-corrosion cracking. Always
Life (Cycles) ensure clean, smooth working sur-
faces free from contaminants.

58 Professional BoatBuilder
 this reason the United Kingdom
Ministry of Defense accepts only O-
and fabrication. And the cost of need or the expense. This is QA at temper alloys, which are the softest
repairing a finished and flawed ves- its best. By learning and adhering and most ductile. A typical value of
sel can equal a large percentage of to similar rules and standard prac- proof stress of 5083 alloy in the O
the initial build cost. tices, and recognizing that steel and temper is nominally 120–140 MPa
In steel construction, applying the aluminum require different approaches (Fig. 5). Strain-hardening the same
correct filler wire or painting the bare that should not be universally 5083 alloy to a temper of H2 increases
metal surface to prevent obvious cor- applied to both, the same high this proof stress to roughly 240 MPa.
rosion have become standard best quality can be achieved in alumi- That sounds good, as the allowable
practices, so ingrained and accepted num construction. design stress limit has increased. But
than no one questions either the —John Kecsmar closer inspection reveals that the
same strain-hardening has reduced
the margin from yield to ultimate
Figure 5-1. Typical Properties of 5083 Alloy at Different Tempers tensile strength (UTS), or failure.
Shaping should be performed in the Simple processes such as rolling and Fig. 6 illustrates that strain-
O temper (annealed) or H111 (another bending will work-harden the alloy, hardening each temper also increases
temper designation) alloys to avoid which in turn reduces its capacity for the UTS, but at what cost? Thus, the
excessive localized strain-hardening. deformation and load-carrying. For UTS of O temper is approximately
300–330 MPa, whereas in H9 temper,
Figure 5. Typical Properties of 5083 Alloy at Different Tempers the UTS is 420. As strain hardening
increases, the percentage over the
base yield proof stress the temper can
Ultimate Tensile Strength carry, decreases. It ranges from 107%
~330 for O temper, down to a paltry 14% in
the H9 temper.
H2 temper This means that if the alloy is over-
240 loaded from its design-allowable
Stress (MPa)

stress in O temper, the amount of

reserve capacity is 107%, or double
O temper the design limit—an inbuilt safety fac-
140
tor of 2. But in the H9 temper, for
example, the margin from yield to
overload is so small that an overload
can cause instant failure of the struc-
tural member.
In addition, O temper is easier to
bend for hull plating; and the result-
Figure 5-2. Amount of Available “Overload” to Base Metal ing temper—after factoring in strain
Proof Stress Limit of 5083 Alloy Strain hardening from rolling—leaves suffi-
cient overload capacity in the finished
hull components. With O temper, it’s
likely that simple rolling of hull plat-
Figure 6. Amount of Available “Overload” to Base Metal ing can increase temper to, say, H116.
Proof Stress Limit of 5083 Alloy But an already strain-hardened alloy,
say H321, is significantly more diffi-
cult to roll, and once rolled is strain-
% of Load Over the Base Proof Stress

120 hardened possibly to a temper of H2

or more, depending on the curva-
100 ture’s complexity. This leaves little
capacity for deformation in the event
80 of overloading.
Some old-school fabricators will
60 apply the heat method common in
steel fabrication to straighten a buck-
40 led or bent aluminum plate. This is
poor practice. Heat can significantly
20 affect aluminum’s proof strength.
Even if the metal has been perfectly
0 handled, cut, shaped, and fixed into
0 H2 H4 H6 H8 H9 position, it must still be welded. This
Temper process alters aluminum’s fatigue
strength, sometimes dramatically if

June/July 2012 59 

 Figure 6. S-N Curves of 5000 Series Weldments

Figure 7. S-N Curves of 5000-Series Weldments correct procedures are not followed.
Welding reduces fatigue properties
even further than those shown in
160
Figs. 2 and 4. When aluminum is
140 immersed in a corrosive environment
like seawater, its fatigue life is
120
reduced yet again, as shown in Fig. 7.
Stress (MPa)

100 A welded joint does not have to be

immersed in seawater for its fatigue
80
life to be reduced. If it is wetted just
60 once with seawater in a bilge or
In air unventilated compartment, and then
40
dries, the effect is the same: the salt
20 In seawater deposits are deliquescent, meaning
0 that in damp conditions the salt crys-
Figure 7. Cutouts
100E+04 in Main Frame
100E+05 Web
100E+06 100E+07 100E+08 100E+09 tals attract water in the air to form a
highly saline drop of liquid, probably
Endurance (cycles) more corrosive than seawater.
Fig. 7 illustrates that when alumi-
num is welded, the reduction in
Figure 8. Cutouts in Main Frame Web fatigue strength “in air” is approxi-
mately 50%, and when that weld is
immersed in seawater, reduction can
Transverse frame R10 be as much as 90% of the static
R10 (0.4")
(0.4") unwelded strength. So, static design
limits appear to be rather meaningless
for a welded aluminum boat that is
constantly exposed to or immersed in
seawater. To avert failures, designers
must carefully position welded joints,
Stringer and builders must practice compre-
hensive quality control.
63
R15 Frame Design
(0.6") Proper structural analysis is always
necessary in planning a complex
structure. Incorrectly identifying load
Hull plate paths in connecting structure can
cause serious problems in the framing
for aluminum boats. When a typical
stringer/longitudinal passes through a
main transverse frame, cutouts are
Figure 9. Poor Weld Return required so both structural members
can be continuous. The size, shape,
and connection detail of these cutouts
requires care and an understanding of
the true forces the frames will be sub-
jected to. In Fig. 8, a typical arrange-
ment shows an angle bar stringer,
which could just as easily be a T or a
bulb bar.
In the connection between the web
of the transverse frame and the
web of the longitudinal, seen in sec-
tion, there is a 0.6" (15mm) radius.
This prevents the weld of the trans-
verse frame web (left to right on the
page) from coming in contact with
the weld of the web of the stringer.
That’s important for two reasons:
first, to avoid locating a weld over a
weld (more on this later); second,
the weld in the frame web carries a

60 Professional BoatBuilder
load transversely, and the web of the Figure 10. Good Weld Return
stringer carries a load longitudinally.
When these two welds join, the weld
at the intersection is being “pulled”
in two directions. If a load in each
weld has a unit value of 1, then
owing to the connection, the result-
ing maximum load is the square root
of the triangle of forces: √2, or 1.41.
Thus, the weld at the intersection has
an increase in stress of 41% over “as
designed” stress. Overlay an increase
of 41% in expected stress onto Fig. 7,
and the resulting weakness and
likely failure is apparent, commonly
called a biaxial stress riser. (A triaxial
is a joint in three directions). The
size of this radius needs to ensure
that the two welds don’t touch.
The weld on the frame web must
also return around and be continu- Figure 11. Undressed Weld Stops
ous. If the cutout is an odd shape/
size, this too can increase the local
stress as a discontinuity.
An example of a poor weld return
is shown in Fig. 9. Weld returns are
important because as the shear load
in the weld runs out, the load has
nothing to pass into. (FEA can estab-
lish the structural load path to illus-
trate the phenomenon.) But more
importantly, the end of a weld
always has too much heat without a
runoff plate and creates a heat sink
as the weld solidifies, and eventually
leaves a classic “crater,” or start-stop,
crack. This is visible in the poor joint
shown in Fig. 9, as indicated by the designer must ensure that the shear made crack waiting to propagate.
circle. The arrow is pointing at the and bending loads are not excessive Another common site for cracks is
heat sink and small pore, a micro- in any loading conditions in a sniped- the intersection of the flange of a lon-
scopic flaw, which eventually end structural member. gitudinal and the web of a transverse
becomes a crack that propagates frame. You can see in Fig. 8 that the
under repeated loading. This type of Weld Flaws flange of the angle bar has a defined
flaw is also called a hot crack. Cracks in welds are the most com- gap between it and the frame web.
Shown in Fig. 10, a good return at mon fatigue failure. They are generally Some designers and yards leave no
the ends of a joint will prevent the the result of poor welding practice and gap, thinking that for a really stiff
heat sink, thus eliminating the flaw, not performing the standard checks to joint it’s often best to weld them
and will provide a direct path around maintain quality. One of the most together. This is incorrect. The result-
the joint to distribute the shear load common is a simple start-stop crack ing joint is very stiff; however, the
when the vessel is in service. initiated when too much heat is con- problem that manifests in service is
If stringers are highly loaded, centrated in one location for too long, similar to, but worse than, the biaxial
sniped ends will also exacerbate or not long enough to fully penetrate. stress riser on the lower-radius web
the fabrication problem. A snipe is the Considered to be too time consuming, connection, discussed above. The
common method of finishing a struc- good welding practices are too often weld of the frame web is carrying
tural member if an end doesn’t attach abandoned when a builder is under shear transversely, and the flange in
to another member. For example, the pressure of production. A classic the longitudinal direction carries ten-
when an angle bar stops short of a example is when a welder makes a sile load. The shear strength of alumi-
frame and it’s not possible to attach long run and neglects to grind back num is much less than its tensile
a bracket to the frame, the bar is cut the stop to prevent a void (Fig. 11). If strength by a factor of √3; therefore,
back at an angle of 30°. This gradu- the weld has gone cool or cold, a new the allowable stress that the weld can
ally reduces the change of section, weld run on top, or if performed too take is reduced by 58% (1/√3)—the
thus lessening any stress concentra- quickly, will rarely fuse deep into the difference between the tensile and
tion from loads in the member. The root and so leaves a void—a ready- shear load capacity. Since the stringer

June/July 2012 61 

 Figure 12. Improper Joint Weld is constantly subjected to wave loads,
it flexes with each passing wave in
this extreme fatigue environment. As
a result, the weld will crack, as shown
in Fig. 12. Again, if the designer over-
laid the calculated stress, with the √3
increase at this joint, onto the graph
in Fig. 7, the implications for longev-
ity are obvious. (While Fig. 7 shows a
drop in strength at fatigued welds of
roughly 90% in 5000-series aluminum,
the results would be slightly worse in
Figure 11. Excessive Gap Filling 6000 series.)
Poor fit-up is another common
problem. If the plate is incorrectly
Courtesy Lloyd’s Register

sized, a large gap will result when it’s

fitted up to the hull. Far too often, the
welder will gap-fill to make it fit. A
classic example of this is shown in
Fig. 13, a cross section of the connec-
tion of a stern tube passing through a
waterjet duct, exhibiting a large void.
Figure 13. Excessive Gap Filling The stern tube must be fitted to exact
tolerances to ensure the correct gap
for welding. Though the drawing
called for just three weld beads, or
passes—a root and two caps—numerous
beads are visible on the finished and
Weld toe failed structure. This excessive gap-
filling led to the inevitable lack of
penetration (LOP) and lack of fusion
LOP/LOF void (LOF). Inadequate penetration means
the weld pool does not reach the
weld root, and therefore a root gap
Crack
remains. Finally, where the minimum
Weld cross-sectional area and heat-affected
zone (HAZ) coincide, there’s the
resulting fatigue crack at the weld toe
Figure 12. Lack of Penetration/Lack of Fusion (the line where the base metal meets
the weld metal on the surface). The
stern tube shown in Fig. 13 failed
within a few months of going into
Figure 14. Lack of Penetration/Lack of Fusion
service. Despite being a small joint, it
was a very costly repair, because the
vessel had to be put in dry dock.
300 Poor training or just a fabricator’s
momentary lapse in attention necessi-
Sound
Fatigue Strength (MPa)

tated this expensive correction.

200 welds
0.06" Apart from being extremely
0.2"
(5.3mm) (1.5mm) unsightly, LOP/LOF seriously impacts
the fatigue life of a welded joint, as
illustrated in Fig. 14. That graph high-
RR
Size of lights several other welding defects as
100 imperfection well. As the void size increases, the
and life range
number of cycles to failure decreases
0.15" RI 0.02" significantly, as does the correspond-
(3.8mm) (0.5mm) ing fatigue strength with or without
50 reinforcements. These unfused inter-
103 104 105 106 107 faces between filler metal and base
metal or between different layers of
Cycles to Failure
the filler material—characterized as
D. Kosteas

RR = reinforcement removed RI = reinforcement intact fusion defects, or LOP and LOF—are

difficult to detect with non-destructive

62 Professional BoatBuilder
 Figure 13. Linear Misalignment

testing such as dye penetrant. Figure 15. Linear Misalignment

Poor fit-up affects alignment as well
as gap filling. Misalignment can cause 330
localized stress concentrations
because the resulting weld is over-
310

Stress (MPa)
sized, and also cause localized sec- e t
ondary bending stresses because
of discontinuity between butted 290
plates. The combination of these

D. Kosteas and S.J. Maddox

localized effects exposes the weld toe AIMgSi1/S-AISi5
270
to higher-than-expected stress (see t = 0.10" to 0.13" (2.5mm to 3.2mm)
Fig. 15).
The increase in secondary stresses 250
0 10 20 30 40 50 60 70 80 90 100
is related to the geometric relationship
of plate thickness and distance of mis- Relative Eccentricity e/t(%)
alignment. We can calculate the
increase of stress at the weld toes in
the plate by the simple relationship: Figure 16. Crack in Oversized Weld

sN + sM = sN . (1 + 3.e/t)
Figure 13. Linear Misalignment
Where:
sN is axial stress 330
t is plate thickness
e is eccentricity
310 t
Stress (MPa)

e
The term in parentheses is the
stress-magnification factor, Km. So if, 290
for example, you have an axial mis-
alignment of just 10% of the plate AIMgSi1/S-AISi5
270
thickness (e/t = 0.1), Km = 1.3. A sim- t = 0.10" to 0.13" (2.5mm to 3.2mm)
ilar increase in stress results from a 1°
angular misalignment, which means 250
0 10 20 30 40 50 60 70 80 90 100
that if the plate is off by 10% and the
angle of the joint is also rotated by Relative Eccentricity e/t(%)
just 1°, these slight imperfections
combine for Km = 1 + 0.3 + 0.3 = 1.6.
That’s a 60% increase of stress at the
weld toe.
Fig. 16 shows the result of attempt-
ing to overcome the gap where a
frame rider butts into another rider
running transversely. To fill the exces-
Figure 17. Bracket-Installation Errors
sive gap, the welder “buttered” each
edge with two passes, which cooled
and locked in stress. The weld was
finished by filling the middle with
additional weld beads achieving no
real penetration. Aluminum has
approximately five times the thermal
conductivity, and twice the rate of
thermal expansion, of steel, so the
heat from welding travels farther and
faster than in steel. The crack is
shown by the circles on either side of
the rider, and the arrows indicate the
hairline crack that has surfaced across
the weld. It cracked because the over-
sized volume of weld could not cool
evenly across the whole joint, which
caused internal thermal stress gradi-
ents. Thus, it cracked as it cooled.
Fig. 17 illustrates several bracket­-

June/July 2012 63 

 Figure 16. Rollover (Overfill) Toe Angle

Figure 18. Rollover (Overfill) Toe Angle installation errors. The crack circled
in red is the most obvious one. It was
120 caused by the lack of a proper weld
return, which led to too much heat at
100 the ends as the welder stopped,
Stress Range (MPa)

80 which in turn pulled the joint as it

cooled. To the left of the crack, the
60 welder gap-filled to overcome bad
α toe
Alloy NP5/6 fit-up. You also can just see a hard
40 R=0 saw cut, indicated by the arrow, in
t = 0.37" (9.4mm) the bracket material. This hard edge
20
will also crack.
0 Gaps and misalignment aside, over-
100°
Figure 17. Effects of120°
Weld Dressing 140°
on Fatigue 160° 180° sized or misshapen welds on their
Toe Angle α own are sources of local discontinu-
ity. The stress in a typical weld varies
from the nominal stress—what you
Figure 19. Effects of Weld Dressing on Fatigue expect the weld to carry—to a much
higher peak at the weld toe. How this
affects the fatigue life is shown in
Fig. 18. Within a window of likely
250
weld profiles, the lower the toe angle
base metal (meaning the more rollover), the
200 lower the load the weld can carry. As
Fatigue Strength (MPa)
Aluminum and the Sea/Pechiney Rhenalu Handbook

the allowable stress is lower, so too

is the fatigue life.
150 Rollover (overfill) If we look at the profile of a butt
dressed flush
weld as a discontinuity, it shortens the
100 fatigue life of a structure. Fabricators
can minimize the effect by dressing
Butt weld the weld bead, which reshapes the weld
50 as welded toe to eliminate rollover. The curves
5000-series alloy
shown in Fig. 19 clearly illustrate that
0 18. Effect of Repairs on Grain Size an as-welded joint, which shows the
Figure weld bead profile untouched, has a
103 4 6 104 4 6 105 4 6 106 2 5 107 2 5 108
much lower load-carrying capacity,
Cycles to Failure and thus a shorter fatigue life, than a
dressed weld or the virgin plate.
Welding over a weld, and rewelding
Figure 20. Effect of Repairs on Grain Size existing welds owing to mistakes and/
or replacing plate, are far-too-common
95 poor practices. When subjected to a
typical UTS bend test, these joints
90 show undesirable effects on the grain
Root pass size. Fig. 20 shows an increase in
85 grain size of 33% by the fourth repair.
Face pass Hardness is proportional to yield
Grain Size (μm)

80 strength, so as grain size increases,

hardness decreases, which in turn
75
reduces strength.
70 The effect of multiple welding
repairs on mechanical strength is
S. Katsas, J. Nikolaou, and G. Papadimitriou

65 quantified in Fig. 21. As weld repairs

increase, UTS consistently falls below
60 accepted standards.
Weld temperatures in the heat-
55 affected zone typically range from
932°F to 1,112°F (500°C to 600°C). At
50
1 2 3 4 5
these temperatures, fine particles
0
(as welded) called intermetallics—such as
Number of Repairs Al6(MnFe) or Al6Mn and Al3Fe—
precipitate from the alloy, and are no
longer soluble once the weld cools.

64 Professional BoatBuilder
 The increase in grain growth shown Figure 21. Effects of Welding Repairs on Mechanical Strength
in Fig. 20 is consistent with the
increase in intermetallics and porosity— 290
sites of crack initiation—caused by
repeated exposure to welding temper- 280 ASM Handbook safety limit

Ultimate Tensile Strength

atures. Welding over a weld increases 270
the grain size and the number of
intermetallics inside the alloy with 260
each thermal cycle. This has the effect
250
of greatly reducing strength and the
fatigue life of the welded joint, and 240
expanding the weakened area around
a weld. The specific influence on 230
fatigue life owing to the increase of 220
intermetallics can be as much as a 30%
reduction, and porosity alone can 210
reduce the fatigue life by as much as 200
200 times, with increasing flaw size 0
(Fig. 22). In repairs, the welder should Figure 21. Effects of Porosity 1
(as welded)
2 3 4
Number of Repairs
always cut out the heat-affected zone
to avoid the cumulative degradation of
the metal from repeated welding.

Solutions Figure 22. Effects of Porosity on Fatigue Life

In repairing the original failure
106
in our waterjet-powered vessel, we Butt weld with rollover
identified the fabrication faults that dressed flush
initiated fatigue-related failures in its 105
complex stern structure. With FEA we Factor of ~200
Cycles to Fracture

found the locations of the load paths 104 on life

that the original design and fabrica-
tion had failed to identify as sources 103
of potential failures that could result
in cracking (Fig. 23). By applying
best practices in aluminum to the next 102
Respective increase of net section stress
generation of waterjet-powered ves- in the order of 50%
sels, we have avoided those faults in 10 1

fabrication.
We introduced more transverse 100
frames in the detail design, so we 0 8 16 24 32
could then reduce or completely elim- Porosity in % of Fracture Area
inate longitudinals in the affected
region. This provided easy access,
so the welder could make one con- Figure 23. Load Paths
tinuous weld, thereby limiting the
start-stops that introduce so much
potential for cracking. With no longi-
tudinals or need for cutouts through
structural members, all the connections
were smoothed, minimizing welding
discontinuities.
During fabrication, builders now
roll the plate with more care and
attention, and along the grain rather
than across the grain. Where feasi-
ble, we introduced post-weld treat-
ment to dress welds, reducing the
stress concentrations at the weld
toes (Fig. 24), which has been
shown to improve the fatigue life.
The application of the new waterjet
structure design and fabrication philoso-
phies were fully adopted on a class of

June/July 2012 65 

Figure 24. Post-weld Treatment of Flange-Welded Joint of Jet Duct

Figure 24. Post-Weld Treatment of Flange-Welded Figure 25. Waterjet Duct

Joint of Jet Duct

0.4" (10mm) Back face flange

machining allowance
1" Holes drilled in
(25mm) conjunction with
KaMeWa flange

ø760

ø14
R8

0.04" 20.0°
(1mm) min

147.6' (45m), 200-metric-ton, 45-knot vessels. Twenty-three

metric tons of water per second pass through the waterjet
Depth of grinding duct shown in Fig. 25. To the best of my knowledge, the
to be 0.02" (0.5mm) boats have not experienced a single structural failure
below bottom of any
visible weld of the transom around the waterjets in their 17 years of
service of 12–16 hour days, seven days a week. The same is
true of the smaller repaired waterjet structure highlighted
at the beginning of the article.
0.05" This evidence justifies the close attention to detail that took
(1.25mm)
place during design and fabrication to ensure that a
structure will be trouble-free for its lifetime. In all the waterjet-
Scale 1:1
powered vessels I have designed during the past 17 years, I

66 Professional BoatBuilder
have found that establishing correct
procedures and sticking to them are
the keys to a long-lived and warranty-
free vessel.
About the Author: John Kecsmar
formed the marine consultancy com-
pany Ad Hoc Marine Designs Ltd with
Nigel Warren in 2005, after spending
nearly 20 years together at FBM
Babcock Marine, in Newport, United
Kingdom. John is on Lloyd’s Register
Technical Committee, RINA’s High
Speed technical committee, MCA’s
High Speed Advisory Group, and
SNAME’s O-50 Maritime Quality
Culture Group. He has designed high-
speed aluminum vessels such as
patrol boats, fast ferries, SWATHs, and
crew boats for more than 20 years
and is the author of many technical
papers on high-speed design, structural
design, and fatigue. He lives in Japan.
John dedicates this article to his friend
and mentor, Nigel Warren, who died
during its preparation. John writes
that Nigel was very generous with his
encyclopedic knowledge of boats, and
is sadly missed.
Further Reading
For in-depth study of fatigue failures and aluminum boat construction, we
include a list of the author’s technical papers and other sources he credits
for this article.
Katsas, S., J. Nikolaou, G. Papadimitriou. “Microstructural changes accompany-
ing repair welding in 5XXX aluminum alloys and their effect on the mechanical
properties,” Materials & Design, Vol. 27, Issue 10, Elsevier, 2006.
Kecsmar, Fyfe, Hawkins, Shenoi, Price, Temarel, Read. “Fatigue Performance of
Welded Aluminum Tee Connections,” FAST 1997.
Kecsmar, J. “Guidance on How Weld Quality Influences the Fatigue Life of
Welded Aluminum Structures,” JASNAOE Conf., Vol. 3, Nov. 2006.
Kecsmar, J., R.A. Shenoi. “Some Notes on the Influence of Manufacturing on the
Fatigue Life of Welded Aluminum Structures,” Journal of Ship Production,
Vol. 20, No. 3, SNAME, August 2004.
Kecsmar, J., N. Warren, J. Moore. “The evolution of advanced SLICE®
Technology adapted to satisfy the HSC Code and commercial requirements,”
RINA H.S.A.M.V. Conference, Shanghai, China, April 6, 2007.
Kecsmar, J., N. Warren, N. Sims. “Waterjet Propulsion: A Shipbuilder’s View,”
RINA Conference, London, U.K., December 1–2, 1994.
Kosteas, D. “Fatigue Behaviour and Analysis,” Lecture 2401, Advanced Level,
Training in Aluminum Application Technologies, European Aluminum
Association, 1994.
Maddox, S.J. Fatigue Strength of Welded Structures. Abington Publishing, 2nd
Ed, 1991, ISBN-10 1 85573 013 8.
Panico, J.R. “The effect of multiple weld repairs on the fatigue resistance of
welded aluminum alloy 5083-O,” NTIS, Dec. 1979.
Pechiney Rhenalu Handbook. Aluminum and the Sea, 1993.
Polezhaeva, H., M. Malinowski. “Fatigue Strength of Aluminum Structural Details
of Special Service Craft,” FAST 2001, Southampton, U.K.
June/July 2012 67