Journal of Constructional Steel Research 172 (2020) 106200

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Fatigue crack repair in welded structures via tungsten inert gas remelting
and high frequency mechanical impact
Hassan Al-Karawi a,⁎,1, R.U. Franz von Bock und Polach b, Mohammad Al-Emrani a
a
Department of Architecture and Civil Engineering, Chalmers University of Technology, Sweden
b
Institute for Ship Structural Design and Analysis, Hamburg University of Technology, Germany

a r t i c l e i n f o a b s t r a c t

Article history: Rehabilitation of welded structures has gained increasing attention lately. This paper aims at investigating the ef-
Received 23 April 2020 ficiency of Tungsten Inert Gas (TIG)-remelting and TIG-remelting followed by High Frequency Mechanical Impact
Received in revised form 30 May 2020 treatment (TIG-HFMI) in fatigue life extension. Fatigue tests were carried out on as-welded and cracked speci-
Accepted 2 June 2020
mens after treatment. The lives of the treated specimens increased remarkably by the two methods (TIG and
Available online 29 June 2020
TIG-HFMI). Many of the treated specimens ran-out after 10 million cycles of loading and failed at the clamping
Keywords:
location when tested at a higher stress range. The improvement in compressive residual stresses, hardness values
High frequency mechanical impact and weld toe radii were the reasons behind the life extension. These factors were used for fatigue life estimation
TIG remelting in as-welded and TIG-treated specimens using the base metal S\\N curve. Moreover, the test results together
Crack repair with results from previous tests in the literature demonstrated that these methods can be useful for crack
Crack retrofitting retrofitting as for new structures.
Life extension © 2020 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://
Pre-fatigue creativecommons.org/licenses/by/4.0/).

1. Introduction the local stress concentration. Several studies established the benefit of
TIG-remelting in enhancing the fatigue strength of new structures
One of the major challenges facing the durability of steel structures is [4–11]. Pedersen et al. concluded that TIG-remelting is the best studied
the fatigue of their welded details due to cyclic loading. These details are treatment method for implementation in mass production. Lieurade
more prone to fatigue than other parts of the structure because of their et al. found that treating butt and cruciform welded joints with TIG-
higher susceptibility to cracks. Fatigue in welds can be traced back to melting resulted in fatigue strength improvement of steels with differ-
three reasons: the tensile welding residual stress, the high stress con- ent qualities [12]. Moreover, Lixing et al. studied the performance of
centration at the toe and the presence of different forms of weld defects TIG-treated joints under constant and variable amplitude loadings, it
(e.g. undercut) [1]. Several post-weld treatment methods have been was concluded that the improvement in fatigue strength is approxi-
proposed in the literature to extend the life of the welds and increase mately the same under both loading conditions [13].
their fatigue strength. These methods can be branched into two Few research effort projects were conducted to study the efficiency
categories based on their effect:weld geometry improvers (such as of TIG-remelting in fatigue crack repair. [14–17]. TIG remelting effect
TIG-remelting or burr-grinding) which mainly aim to reduce the stress on cracked T-joint was studied by Ramalho et al. and it was found that
concentration factor, and residual stress improvers (such as peening or the treatment is capable of extending the fatigue life by a factor of 2.5
stress relief) which aim to eliminate the tensile residual stress or induce [15]. Besides, the effect of crack size in cover plate details treated by
compressive residual stress. The International Institute of Welding TIG-remelting was studied by Fisher et.al [16,18]; they found that cracks
(IIW) reported a 30% increase in fatigue strength after TIG-remelting initiated from weld root which demonstrates the significant enhance-
without altering the slope of the S\\N curve [2]. On the other hand, ment in fatigue strength of the weld toe. In additions, TIG remelting
peening was reported to increase the fatigue strength furthermore was found to be the best of the three studied methods. Baumgartner
and reduce the slope of the S\\N curve of the detail [3]. et al. concluded that TIG-treatment can trigger crack initiation in differ-
TIG remelting aims to eliminate the weld flaws through fusing the ent location such as base metal and weld roots [19]. Miki et.al concluded
material at the weld toe by the local heat input. In addition, it provides -after studying both transverse and longitudinal attachment details-
smooth transition between the weld and the main plate which reduces that the life extension depends on the crack depth before treatment
and TIG fusion depth [14].
⁎ Corresponding author.
High Frequency Mechanical Impact (HFMI) is one of the newest and
E-mail address: [email protected] (H. Al-Karawi). most promising treatment method. This method is given in the litera-
1
Full postal address: Västra Andersgårdsgatan 15A, Lgh 1102, 41,715, Göteborg, Sweden ture by different names: Hammer peening, Ultrasonic impact treatment,

https://doi.org/10.1016/j.jcsr.2020.106200
0143-974X/© 2020 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
2 H. Al-Karawi et al. / Journal of Constructional Steel Research 172 (2020) 106200

Table 1
Nomenclature The chemical composition and the mechanical properties of S355 and C6LF filler [37].

Δσ applied stress range Chemical composition

ΔσEQV equivalent stress range C Si Mn S P

Δσar fully reversed stress range S355 % 0.23 0.05 1.6 0.05 0.05
C6LF % 0.03 0.73 1.51 0.01 0.01
Δσl local notch stress range
Δσg nominal applied stress range Mechanical properties
σm, g nominal mean stress
Yield strength MPa Ultimate strength MPa Elongation %
σm, l local notch mean stress
σu, g tensile strength of S355 steel S355 355 575 22
C6LF 462 538 30
σu, l local notch tensile strength
σRS residual stress at the weld toe
σclamp clamping stress at the weld toe
Kt stress concentration factor Table 2
A, B Basquin's equations constancts Welding parameters.
Nf, calc calculated number of cycles to failure Run Current A Voltage V Speed mm/s Heat input KJ/mm
Nf, exp obtained number of cycles to failure from tests
1 240 28.3 5.8 0.9
ϕ distortion angle 2 235 28.3 5.6 0.95
m slope of the S\\N curve
Nf number of cycles

Abbreviations available knowledge on the use of TIG-HFMI treatment in repairing fa-

F weld toe failure tigue cracks. Therefore, this paper presents some fatigue test results
CF failure at the clamping position and several experimental investigations on the efficiency of these two
RO run-out methods in fatigue crack repair and shed the light on the factors behind
TIG tungsten inert gas life extension. Moreover, a simple calculation tool based on Basquin's
HFMI high frequency mechanical impact treatment Equation is used to incorporate these factors to predict the fatigue life
IIW international institute of welding of the treated specimens. A special light will be thrown on comparing
FAT fatigue strength class the strength of the treated new and cracked structures.

2. Testing program
Ultrasonic needle peening and HiFIT. HFMI has a triple beneficial effect
on welds; it induces compressive residual stress at the weld toe and 2.1. Test specimens and materials
contributes in local material hardening. In additions, it reduces the
stress concentration locally [20]. HFMI has been used extensively in In light of the literature review, fatigue testing program was set up
the literature for repairing cracked structures [21–31]. Unlike TIG- consisting of 27 transverse attachment specimens. The dimensions of
remelting, HFMI causes crack closure and does not remove the crack the specimen are given in Fig. 1. The specimens were made of mild
completely [20]. Nonetheless, combining TIG-remelting with HFMI- steel S355. The welding filler material was made of core weld C6LF
treatment causes crack removal and introduces compressive residual (metal cored wire). The chemical composition and the mechanical
stress. This combination was found to be superior to the individual properties of the steel and the filler material are listed in Table 1. The
TIG or HFMI treatment of welds [9,32,33]. However, TIG-HFMI treat- properties and the compositions were obtained from the materials cer-
ment has not -to the best of the author's knowledge- been investigated tificates [35,36].
as a crack retrofitting method in the literature.
There is no consensus in the literature on the allowable crack size
2.2. Welding and fabrication
which can be removed by TIG-remelting; this is attributed to the vari-
ability in crack detection methods, fusion depths, detail types, heat in-
Metal active gas welding with a Mison 18 shielding gas were used to
puts and applied loading levels. However, A common conclusion was
join the attachment and the base plate. The attachments were prepared
found in the literature about fusion depth which should be always
greater than existing crack depth to give satisfactory results [14]. This
necessitates the use of relatively precise crack detection methods and
ensuring the fusion depth regularity over the treated area [34].
TIG remelting requires highly skilled labor which explains its scar-
city in the literature for crack retrofitting. In addition, there is no

Fig. 2. Welding procedures: 1: Machining the stiffener. 2: Tack welding. 3: Root welding.
Fig. 1. The studied specimens geometry (dimensions in mm). 4: Cap welding.
 H. Al-Karawi et al. / Journal of Constructional Steel Research 172 (2020) 106200 3

Hamburg University of Technology (TUHH). The positive stress ratio of

0.29 was chosen in order to simulate the real condition in a bridges
where the sustained dead load is taken into account. The testing rig
transferred the axial force into the specimens by the means of the hy-
draulic clamp as shown in Fig. 4. The applied force is measured on the
counter side of the force excitation. The specimens were grouped into
four series as shown in Table 3. Noticeably, the specimens in series C
and D - which are the core of the testing- were tested twice, the first
to generate fatigue cracks (Prefatiguing stage), and the latter to investi-
gate the efficiency of the treatments after Prefatiguing stage.

2.4. Supportive investigations

Several auxiliary investigations were conducted to get deeper in-

sight on the treatment's efficiency. The in-depth residual stress distribu-
tion was investigated by the hole drilling method. The investigation was
conducted on different specimens in as-welded and treated conditions.
Fig. 3. Strain gauge configuration on test specimen. Besides, a 3D laser scanner was employed to explore the changes in
local geometry at the weld toe. Mainly two parameters were studied:
weld toe radius and undercut depth. Moreover, Vicker's hardness tester
by machining their ends with 30∘ angle and 0 mm nose to provide better with a test load of 3 kg weight was used. Hardness mapping was
accessibility and reduce the lack of fusion risk. Subsequently, the attach- performed with a spacing of 0.5 mm in the horizontal and vertical direc-
ment was tack welded to the base plate, with a gap of 3 mm in order to tions. Thus, a metallographic inspection was conducted to ensure the
join them during the subsequent welding runs. Afterward, two subse- regularity and the adequacy of the fusion depth and to visualize the
quent weld runs (root and cap weld) were carried out. The parameters dominant microstructure before and after treatment. Some specimens
of the weld runs are listed in Table 2, and the welding procedure is were cut and sliced perpendicular to the welding line. The surfaces of
shown in Fig. 2. the slices were processed by polishing and Nital-etching with 2% to be
Before fatigue testing, the specimens were instrumented with sev- prepared for optical microscopy.
eral strain gauges at their weld toe to measure the strain drop due to Both destructive and non-destructive methods were used to verify
crack initiation as in [15,22–25]. The strain gauges have an overall and calibrate the use of strain drop measured by attached strain gauges
length and width of 6 and 3 mm respectively. The center of the gauge for crack detection for the specimens in series B. Hence, a red dye pen-
is placed 3 mm off the weld line. The edge to edge spacing between etrant was applied to stain the crack surface. Meanwhile, the specimens
two adjacent gauges was set to be of around 10 mm. Moreover, an addi- back walls were scanned continuously by ultrasonic device to identify
tional gauge was placed 40 mm off the weld line to measure the mem- the crack location and position. Finally, some specimens were cut to
brane stresses. The strain gauge configuration with respect to the monitor the crack profile.
attachments is shown in Fig. 3.
2.5. Crack detection
2.3. Fatigue test setup
In addition to the multiple strain gauges, the specimens in series B
Constant amplitude axial fatigue tests were performed with a stress were scanned by an ultrasonic testing device to inspect the crack
ratio of 0.29 and a loading frequency of ≈ 30 Hz at the laboratory of depth and dye penetrant was applied to stain the crack surface.

Fig. 4. L: Schematic figure shows the hydraulic clamp. R: Fatigue testing rig.

Table 3
The specimens series.

Series N.o of spec Function Specimens Test stop criteria

A 12 Constructing the S-N curve for the as-welded 1–10, 12–13 Failure/Run-out
B 3 Crack detection and verification 14–16 50–80% drop in strain
C 7 Studying the life extension by TIG 17–23 Before treatment: 25% drop in strain After treatment: Failure/Run-out
D 5 Studying the life extension by TIG-HFMI 25–29 Before treatment: 25% drop in strain After treatment: Failure/Run-out
4 H. Al-Karawi et al. / Journal of Constructional Steel Research 172 (2020) 106200

Table 4
TIG-remelting parameters.

Voltage V Current A Speed mm/min

In the present study 11.1–12 270–272 100

Recommended by the IIW [3] 12–17 160–250 80–160

Heat input Nozzle diameter Electrode diameter

KJ/mm mm mm

In the present study 1.78–1.96 12 3

Recommended by the 1.0–2.5 10–14 3–4
IIW [3]

of 60–70°; this range was recommended when HFMI is to be used for

Fig. 5. Crack depth, strain drop correlation. crack retrofitting [20].

3. Results
Afterward, the specimens were opened to monitor to visualize the crack
The as-welded specimens in series A were subjected to different
surface. A correlation was found between the strain drop detected by
stress ranges between 118 and 175 MPa. The fatigue test results are
one of the attached strain gauges and the existing crack depth at the
given in Table 5 and summarized in Fig. 7. Burst fracture of the speci-
weld toe as shown in Fig. 5. Consequently, the specimens in series C
mens was set to be the tests abort criterion. However, three specimens
and D- which are the testing program- were tested and the abort crite-
ran-out after being subjected to 10 million of cycles. Therefore, the
rion was set to be 25% drop in any of the attached strain gauge. This per-
stress ranges were increased to achieve failure within reasonable num-
centage would correspond to a crack size between 0.6 mm and 1.2 mm
ber of cycles. Notably, the obtained characteristic fatigue strength (FAT)
as shown in the figure. Fig. 6 shows the strain plots against the number
was found to be 125 MPa which is 4 fatigue class higher than the FAT
of cycles Nf for some specimens in series C and D. Despite the specimens
value for transverse attachment stated in the Eurocode [38]. The fatigue
being identical in shape, dimensions, welding and manufacturing
strength was estimated by the prediction interval method given by [39].
method, the prefatigued lives were widely scattered which might be
It is noteworthy that the run-out tests were excluded from the evalua-
traced back to the intrinsic disparities between the specimens and to
tion of the FAT value.
the difference in crack initiation location with respect to the strain
After treatment, the specimens in series C and D showed significant
gauges.
fatigue strength when tested under 150 MPa. Many specimens ran-out
after 10 million cycles of loading. Consequently, the loading level was
2.6. Crack repair
increased to 180 MPa. Nonetheless, some specimens ran-out again
after 10 million cycles of loading. Subsequently, the specimens were
The welds of the specimens in series C and D were treated by tung-
tested under higher stress levels of 220 or 250 MPa and in all of them fa-
sten electrode to melt the toes vicinity. The application of a relatively
tigue cracks initiated and propagated at the base plate from the
high heat input ensured deep fusion. The electrode was positioned at
clamping location as shown in Fig. 8. Since the specimens were sub-
the weld toe which thereby created larger undercut at the resulted toe
jected to different stress ranges starting from 150 MPa and ending by
position. However, fitting the electrode at the weld toe secured that
220 or 250 MPa, the equivalent stress ranges ΔσEQV were used to present
the position where maximum fusion was reached corresponded to
the results as in Fig. 7. ΔσEQV is calculated as follows:
crack plane. The specimens were clamped from both sides during
remelting and they were air-cooled after treating the first side. Then,
0
1m1
treating the second side was performed. The heat input were halved ∑ nj  Δσ m
j
at the edges to avoid overheating and subsequent inflammation. The Δσ EQV ¼@ A ð1Þ
∑nj
TIG-remelting parameters fall within the recommended ranges as
shown in Table 4. The specimens in series D were then HFMI-treated
by a single 5 mm radius HiFIT indentor. The angle of inclination of the where:
axis of the indenter with respect to the plate surface was in the range Δσj: Applied stress range.
j: Applied stress range index.
nj: Number of cycle at applied stress range Δσj.

Table 5
Fatigue tests results of series A (As-welded).

Test Spec Δσ [MPa] Nf Test Spec Δσ [MPa] Nf

A1a 1 119 10,000,000 RO A4 4 145 1,080,000 F

A1b 1 143 2,150,000 F A5 5 175 757,000 F
A2a 2 114 7,850,000 RO A6 6 170 685,000 F
A2b 2 125 10,400,000 RO A7 7 165 8,590,000 F
A2c 2 140 8,000,000 RO A8 8 160 9,870,000 F
A2d 2 132 2,750,000 RO A9 9 150 1,130,000 F
A2e 2 145 884,000 F A10 10 150 1,600,000 F
A3a 3 132 2,530,000 RO A12 12 150 3,250,000 F
A3b 3 139 3,100,000 RO A13 13 150 2,420,000 F
A3c 3 145 1,080,000 F

Fig. 6. Examples strain gauge measurements. RO: Run-out F: Failure.

 H. Al-Karawi et al. / Journal of Constructional Steel Research 172 (2020) 106200 5

recommendation [2]. This indicates the efficiency of the treatment

methods in crack retrofitting.
Fig. 9L shows the residual stress distribution measured by the hole
drilling method. The measured residual stresses are transversal to the
welding line (i.e. parallel to the loading direction). Several holes were
drilled in each specimen and the average values were documented; all
of them were placed at the same distance of the weld line. Compressive
residual stresses were found at the surface even in as-welded speci-
mens. Following the treatment, two remarkable changes could be ob-
served: The compressive stress at the surface became higher and the
gradient of the curve became shallower. The combined TIG-HFMI treat-
ment induced further compression as a result of the high indentor's im-
pact energy. Analogous results were found for the local hardness as
shown in hardness mapping in Fig. 10.
Vicker's hardness values and the corresponding tensile strength are
extracted at the weld toe and plotted in Fig. 9R. Moreover, a significant
Fig. 7. The fatigue test results of the as-welded and TIG-remelted and TIG-HFMI treated
increase in the weld toe radius was achieved after both of the treatment
specimens.
methods which thereby caused a reduction in stress concentration fac-
tor as shown in Fig. 11. However, TIG-remelting caused an increase in
the undercut depth; this is traced back to the positioning of the TIG elec-
trode right at the weld toe and not according to the IIW recommenda-
tions [3]. The measured toe radii and undercut depth are given in Fig.
12 and Table 7. The table also shows that the scatter of the toe radius be-
came significantly smaller after TIG-remelting.
The regularity and the adequacy of the fusion depth were investi-
gated by a metallographic inspection as mentioned in Section 2.4. In
total, 14 slices of the cross-section, cut orthogonal to the weld, were in-
vestigated by the optical microscope and an average fusion depth was
found to be 2.06 mm with a standard deviation of 0.17 mm. Moreover,
the minimum detected fusion depth was 1.36 mm which is larger than
the existing crack size (0.6–1.2 mm). This indicates full fusion of the
crack surfaces. Moreover, the heat affected zone size was found to be
similar for both the MAG-welding and TIG-remelting despite the differ-
ence in arc size. This was the result of imposing higher heat input in TIG-
remelting than MAG-welding. The inspection showed relatively big
Fig. 8. Fatigue failure in the base metal of the TIG-treated specimen. pores in the middle of the attachment indicating a lack of fusion. How-
ever, no crack was initiated there because of the low stress concentra-
tion in this region. All of the aforementioned observations are shown
m: Presumed slope of the S\\N curve (For as-welded and TIG- in Fig. 13.
treated specimen m = 3, For HFMI-treated specimens m = 5). The microscopic constituents of the specimens before and after TIG
The test results of the specimens in series C and D are given by remelting are shown in Fig. 14. No significant difference was detected
the red and blue solid rectangles respectively in Fig. 7 and the de- between the fusion zone before and after treatment where acicular fer-
tailed results of the test are given in Table 6. The S\\N curves of rite was found to be the main constituent. However, Widmanståtten
these series could not be constructed because they failed at the ferrite was detected after TIG-remelting, while islands of
clamping position. However, the tests showed that all the speci- allotriomorphic ferrite were found in the weld zone before treatment
men failures lay above the design S\\N curve of the TIG-treated which might be attributed to the slower cooling rate of the TIG-
new transverse attachment detail proposed by the IIW remelting in comparison to the welding. Besides, Bainite morphology

Table 6
Fatigue test results of the treated specimens.

Test Spec Status Δσ [MPa] Nf Test Spec Status Δσ [MPa] Nf

A17 17 TIG 150 10,000,000 RO A22 22 TIG 150 10,000,000 RO

B17 17 TIG 180 10,000,000 RO B22 22 TIG 250 334,159 CF
C17 17 TIG 110 10,000,000 RO A23 23 TIG 150 10,000,000 RO
D17 17 TIG 250 2,434,252 CF B23 23 TIG 250 1,405,626 CF
A18 18 TIG 150 10,800,000 RO A25 25 TIG-HFMI 220 10,000,000 CF
B18 18 TIG 180 10,000,000 RO B25 25 TIG-HFMI 250 2,449,821 CF
C18 18 TIG 220 18,000,000 CF A26 26 TIG-HFMI 180 1,490,000 CF
A19 19 TIG 150 200,000,000 RO A27 27 TIG-HFMI 150 10,000,000 RO
B19 19 TIG 180 10,331,885 RO B27 27 TIG-HFMI 220 1,515,187 CF
C19 19 TIG 220 740,910 CF A28 28 TIG-HFMI 150 10,000,000 RO
A20 20 TIG 150 10,000,000 RO B28 28 TIG-HFMI 200 1,419,626 CF
B20 20 TIG 220 2,491,395 CF A29 29 TIG-HFMI 180 5,648,000 RO
A21 21 TIG 150 10,000,000 RO B29 29 TIG-HFMI 220 2,409,850 CF
B21 21 TIG 250 728,582 CF

RO: Run-Out CF: Clamping failure.

6 H. Al-Karawi et al. / Journal of Constructional Steel Research 172 (2020) 106200

Fig. 9. The In-depth residual stress (Parallel to the loading direction) and the local tensile strength distribution of the as-welded and treated specimens.

dominated in the HAZ of the as-welded specimens and it was tempered The relative displacement of different points along the as-welded
after treatment because of the high heat input. The TIG fusion zone specimens was measured so they could be compared to the correspond-
which mainly consists of acicular ferrite replaced the heat affected ing points after treatment as shown in Fig. 15. The TIG-remelting caused
zone in as-welded condition(which consists of bainite) at the weld additional distortions in the longitudinal direction and twisting the
toe. The interlocking nature of the acicular ferrite, together with its specimens in the transverse direction. This caused a divergence of the
small grain size caused a remarkable increase in the local hardness at two curves corresponding to TIG-remelting in Fig. 15. The high distor-
the weld toe as shown in Figs. 9 and 14. tion was caused by the high heat input used to achieve deep fusion.

Fig. 10. The hardness distribution in the vicinity of the weld in: a:as-welded b:TIG-treated c:TIG-HFMI treated d: Vicker's hardness indent sample.
 H. Al-Karawi et al. / Journal of Constructional Steel Research 172 (2020) 106200 7

Fig. 11. The elastic stress concentration factor.

Fig. 13. Microscopic investigation of the weldment after TIG remetling.

On the other hand, TIG-HFMI treated specimens in series D were less
distorted than TIG-treated specimens in series C because of the applied
constraint on the specimen when HFMI treated.
Since the clamping stresses were derived from the strain measure-
ment 40 mm of the welds, the local stress at the weld toe could not be
4. Discussion directly obtained. Therefore, a linear elastic 2D finite element analysis
was conducted using the commercial software ABAQUS/CAE2017 to
The as-welded specimens showed a remarkably long fatigue life as evaluate the local stress due to clamping. In order to eliminate the stress
mentioned in the previous section. This might be explained by the pres- singularity, a toe radius of 1 mm was used in accordance with the IIW
ence of compressive residual stress at the weld toe as shown in Fig. 9. On recommendation regarding the effective notch approach [2]. 4-noded
the other hand, straightening of the distorted specimens upon clamping bi-linear plane stress elements (CP4SR in Abaqus notation) were used
by the test machine would give rise to tensile mean stress at the weld to create the mesh and a convergence study for the mesh size led to
toe which counteracts the effect of compressive residual stress. In light an element size of 0.025 mm in the weld toe vicinity and 2 mm
of that, the relation between the angular distortion and the clamping elsewhere.
stress was studied thoroughly. The clamping stresses were evaluated Several models were created with several specimen's geometries
experimentally by the attached strain gauge 40 mm away from the based on the measured angular distortions. The elasticity modulus and
weld line (i.e. SG 11 shown in Fig. 3). The expected correlation between Poisson's ratio are chosen to be 210 GPa and 0.3 respectively. The
the clamping stress and the angular distortion was found, see Fig. 16. clamping jaws were simulated as solid undeformed bodies approaching

Fig. 12. The weld toe radius and undercut height normal distributions before and after treatment.

Table 7
The geometrical parameters of the tested specimens before and after treatments.

Status AW TIG TIG-HFMI

The geometrical parameter Toe radius Undercut Toe radius Undercut Toe radius Undercut

Mean 0.67 mm 0.004 mm 5.09 mm 0.150 mm 5.39 mm 0.14 mm

Standard deviation 0.31 mm 0.03 mm 0.73 mm 0.10 mm 2.44 mm 0.06 mm
Variation coefficient 0.46 8.74 0.14 0.67 0.45 0.41
Population size 20,660 20,660 150 1581 334 351
8 H. Al-Karawi et al. / Journal of Constructional Steel Research 172 (2020) 106200

Fig. 14. Micrographic picture for the fusion zone (Top) and heat affected zone (Bottom) before and after TIG-remelting.

each other in a displacement control manner (i.e. vertical displacement the specimen got fully straightened and the inclination of the
was predefined). The displacement of the jaws was restricted in the specimen's wings became zero. The model results are shown in Fig. 17.
horizontal direction (ux = 0). Besides, hard normal contact was defined The obtained nominal stresses were extracted at the location of the
between the clamping jaws (i.e. the master surface) and the specimen strain gauges (40 mm of the weld line). The model was validated by
(i.e. the slave surface). Subsequently, the results were extracted when comparing them against the experimentally obtained nominal stresses
and the differences were found to be insignificant as shown in Fig. 16.
Subsequently, the local stress at the notch was obtained along the
weld toe. The average angular distortion in as-welded condition of
0.26 corresponds to local clamping stress of 90 MPa which is thereby
less than the existing compressive residual stresses of −176 MPa.
Despite the long fatigue life obtained for the as-welded specimens,
both treatment methods resulted in significant fatigue lives extension.
That is conspicuous in Fig. 7 where the solid red and blue rectangles
(TIG, TIG-HFMI respectively) are located to the top of the black rectan-
gle (AW). The successful fatigue life extension by the treatment can be
explained by the shallowness of the existing crack and the efficiency
of the treatment in terms of local geometry improvement, residual
stress change, and local material hardening. The combined effects of
these three factors could be grasped by Goodman's mean stress correc-
Fig. 15. The distortion of the specimens before and after treatments. tion given in Eq. (5).

Fig. 16. The correlation between the angular distortion and the clamping stress. Fig. 17. The finite element model for simulating the clamping process.
 H. Al-Karawi et al. / Journal of Constructional Steel Research 172 (2020) 106200 9

Fig. 18. Incorporation of different factors to evaluate the fatigue life locally.

Basquin's Equation is a power-law relationship which describes tensile strength of the material σu, l. Therefore, no predictions were
the linear relationship between the applied stress range (Δσ ar ) made when Δσar > σu, l.
and the number of cycles to failure (N f ) plotted in a log-log
scale, see Eq. (6). Basquin's equation's constants for S355 steel Δσ l ¼ K t  Δσ g ð2Þ
(A and B) were obtained from [37]. Since the mean stresses (i.e.
σ m,l ¼ K t  σ m,g ð3Þ
axial mean stress, residual stress and clamping stress) and the
measured local ultimate strength were taken into account, the
σ m,tot ¼ σ m,l þ σ RS þ σ Clamp ð4Þ
S\\N curve of the base metal could be used instead of separate
S\\N curves for as-welded, TIG-dressed and TIG-HFMI treated de- Δσ l
tails. This concept is illustrated in Fig. 18. Δσ ar ¼ ð5Þ
1− σσu,l
The calculated fatigue life Ng, calc using Eqs. 2, 3, 4, 5 and 6 are com-
pared to the test results for as-welded and treated specimens Nf, exp.
N f ,cal ¼ A  Δσ Bar ð6Þ
See Table 8. The stresses are given in MPa in the table. The calculated
as-welded fatigue life under a stress range of 150 MPa was found to
where Δσar is the fully reversed stress range considering the mean
be around 2.61 million cycles which lies in the interval of the obtained
stress, while Δσl and Δσg are the local and global stress ranges respec-
experimentally determined fatigue lifes (1.11–3.25 million cycles)
tively. The global and local mean stresses due to external loading are de-
under the same stress range. In addition, most of the test results con-
noted in the equations by σm, g and σm, l respectively. The additional
firmed the model predictions for the treated specimens' lives since no
mean stresses due to residual stress σRS and clamping stress σClamp were
toe failure was obtained when the specimens were tested to Nf, exp as
obtained from the distributions shown in Figs. 9 and 16 respectively. Be-
shown in the table. Nonetheless, Basquin's Equation should not be
sides, the local tensile strength at the weld toe and the ultimate strength
used when the fully reversed stress range is greater than the local
of S355 structural steel are given in the equations as σu, l and σu, g

Table 8
Calculated fatigue lives for as-welded and treated specimens under different loading levels.

σRS σu, l σClamp σm, g Δσg Kt σm, l σm, tot Δσl Δσar Nf, cal Nf, exp

AW −176 627 90 136 150 2.150 292 206 323 481 2.61E6 1.11E6–3.25E6
TIG −219 697 198 136 150 1.366 186 165 205 279 1.16E9 1.0E7
TIG-HFMI −293 767 171 136 150 1.360 185 62 204 268 1.79E9 1.0E7
AW −176 627 90 162 180 2.150 348 262 387 665 – –
TIG −219 697 198 162 180 1.366 221 200 246 345 1.08E8 1.0E7
TIG-HFMI −293 767 171 162 180 1.360 222 98 245 281 1.10E9 1.50E6
AW −176 627 90 200 220 2.150 430 334 473 1047 – –
TIG −219 697 198 200 220 1.366 275 252 300 470 3.29E6 –
TIG-HFMI −293 767 171 200 220 1.360 273 150 299 371 4.66E7 1.50E6
AW −176 627 90 227 250 2.150 488 402 537 1499 – –
TIG −219 697 198 227 250 1.366 310 289 341 583 3.01E5 3.30E5
TIG-HFMI −293 767 171 227 250 1.360 308 186 340 450 5.56E6 2.40E6
10 H. Al-Karawi et al. / Journal of Constructional Steel Research 172 (2020) 106200

Fig. 19. Fatigue test results of TIG and TIG-HFMI treated transverse and longitudinal attachments respectively.

respectively. The stress concentration factor at the weld toe Kt were ob- specimens containing relatively deep crack (deeper than 5 mm). This
tained from Fig. 11. indicates the successful application of TIG-remelting in fatigue crack
Several fatigue test results of TIG-treated and TIG-HFMI treated retrofitting.
transverse and longitudinal attachment specimens were collected The combined effect of TIG-dressing followed by HFMI-treatment
from the literature [7,9,14,15,17,32,33,40–42] and are presented in Fig. has - to the best of the authors' knowledge - never been investigated
19. No distinction was made based on the failure position, mean stress, as crack retrofitting technique. However, it was reported in the litera-
stress ratio or plate thickness. The fatigue test results of the treated ture that it can be used to strengthen new welded structures [40].
specimens \\In this paper- are denoted by the black and red triangles Though the tested specimens in this paper did not fail at the weld toe,
in the figure. The characteristic S\\N curve of the as-welded, TIG- their fatigue lives were long as shown in Fig. 19 and lying way above
dressed and HFMI-treated specimens are also shown in the figure. the design curve of the TIG-treatment for this detail. Despite that the
For transverse attachments, more than 98% of the collected data local stress range for the TIG-HFMI treated specimens is lower than
points exhibited longer fatigue life than the characteristic life for new TIG-treated ones as shown in Table 8, no quantitative comparison
TIG-treated attachment (i.e. Un-prefatigued) given in the IIW [2] at could be made between the tested TIG'ed. and TIG-HFMI'ed. specimens
the same stress range. The only two points lying below the line are re- because none of them failed at the weld toe.
ferred to prefatigued specimens containing deep crack at the weld toe
(deeper than 5 mm). Otherwise, the rest of the prefatigued specimens 5. Conclusions
(Including the tested ones in this paper) are lying within the scatter
band of the new ones (i.e. Un-prefatigued ones). The resistance of the Fatigue tests were carried out on S355 structural steel plates with
longitudinal attachments was found to be more scattered because of welded transverse attachment to investigate the performance of two
its dependency on the attachment length. However, the prefatigued post-weld treatment methods (TIG-remelting, TIG-HFMI treatment) in
specimens showed relatively long fatigue lives with an exception for fatigue crack retrofitting. A crack detection methodology was applied
 H. Al-Karawi et al. / Journal of Constructional Steel Research 172 (2020) 106200 11

based on strain drop. Several auxiliary investigations were carried out to [3] P. Haagensen, S. Maddox, IIW recommendations on post weld improvement of steel
and aluminium, IIW Doc 13 (2003)1815–00.
give deeper insight on treatment effects. The following conclusion can [4] P. Haagensen, et al., TIG Dressing of Steel Weldments for Improved Fatigue Perfor-
be drawn: mance, 1979.
[5] L.C. Wu, D.P. Wang, Improve the fatigue performance of welded joints with under-
• The as-welded fatigue life was found to be relatively long which is cuts by TIG dressing treatment, 472, 2012 1300–1304.
[6] H. Mettänen, et al., Fatigue strength assessment of TIG-dressed ultra-high-strength
mainly attributed to the presence of compressive residual stress at steel fillet weld joints using the 4R method, Int. J. Fatigue (2020) 105745.
the weld toe. [7] H.C. Yldrm, Review of fatigue data for welds improved by tungsten inert gas dress-
• No remarkable difference was observed in the microstructure of the ing, Int. J. Fatigue 79 (2015) 36–45.
[8] S. Van Es, M. Kolstein, R. Pijpers, F. Bijlaard, TIG-dressing of high strength butt
fusion zone after welding and re-melting and the same applies to
welded connections–part 2: physical testing and modelling, Procedia Eng. 66
the heat-affected zone. (2013) 126–137.
• Both treatment methods caused a substantial change in the local weld [9] M.M. Pedersen, O.Ø. Mouritsen, M.R. Hansen, J.G. Andersen, J. Wenderby, Compari-
son of post-weld treatment of high-strength steel welded joints in medium cycle fa-
topography where the toe radius increased with a factor of 7.
tigue, Welding World 54 (7–8) (2010) R208–R217.
• Despite the compressive residual stress found in the as-welded state, [10] S. Van Es, M. Kolstein, R. Pijpers, F. Bijlaard, TIG-dressing of high strength steel butt
TIG-remelting caused a further increase in the compressive stresses at welded connections–part 1: weld toe geometry and local hardness, Procedia Eng. 66
the surface and a decrease in the stress distribution gradient. HFMI- (2013) 216–225.
[11] M. Nascimento, H.J.C. Voorwald, J.D.C.P. Filho, Effects of several TIG weld repairs on
treatment induced further compression at the weld toe. the axial fatigue strength of aisi 4130 aeronautical steel-welded joints, Fatigue Fract.
• TIG-remelting caused an increase in the hardness value at the weld Eng. Mater. Struct. 35 (3) (2012) 191–204.
toe since the heat-affected zone was replaced by a TIG fusion zone [12] H. Lieurade, I. Huther, F. Lefebvre, Effect of weld quality and postweld improvement
techniques on the fatigue resistance of extra high strength steels, Welding World 52
with an acicular microstructure. The HFMI treated zone experienced (7–8) (2008) 106–115.
further local hardening because of the cold working effect. [13] L. Huo, D. Wang, Y. Zhang, Investigation of the fatigue behaviour of the welded
• A correlation was found between the clamping stresses and the angu- joints treated by TIG dressing and ultrasonic peening under variable-amplitude
load, Int. J. Fatigue 27 (1) (2005) 95–101.
lar distortions. Besides, a good match was found between the finite el-
[14] C. Miki, T. Mori, S. Tuda, K. Sakamoto, Retrofitting fatigue-cracked joints by TIG arc
ement results and the experimentally measured clamping stresses. remelting, Doboku Gakkai Ronbunshu 1987 (380) (1987) 111–119.
TIG-remelting increased the distortions while HFMI-treatment [15] A.L. Ramalho, J.A. Ferreira, C.A. Branco, Fatigue behaviour of T welded joints rehabil-
reduces them. itated by tungsten inert gas and plasma dressing, Mater. Des. 32 (10) (2011)
4705–4713.
• A significant increases in fatigue lives was obtained by both of the in- [16] J.W. Fisher, A.W. Pense, R.E. Slockbower, H. Hausammann, Retrofitting fatigue dam-
vestigated treatment methods. Thereby, many specimens ran-out aged bridges, Transp. Res. Rec. (664) (1978).
after 10 million cycles when subjected to membrane stress range of [17] H. Al-Karawi, A. Manai, M. Al-Emrani, R. Von Bock, N. Friedrich, J. Hedegård, Fatigue
crack repair by TIG-remelting, The International Conference on Bridge Maintenance,
150 MPa. Safety and Management IABMAS (submitted and accepted), Sapporo, Japan, 2020.
• When the treated specimens were tested at stress level larger than [18] J. Fisher, M. Sullivan, A. Pense, Improving fatigue strength and repairing fatigue
150 MPa, base metal failures from the clamping position were ob- damage, Final Report, NCHRP Project 1974, pp. 12–15.
[19] J. Baumgartner, H.C. Yldrm, Z. Barsoum, Fatigue strength assessment of TIG-dressed
served. This emphasizes the strength of weld toe position after TIG-
welded steel joints by local approaches, Int. J. Fatigue 126 (2019) 72–78.
remelting in case of full crack removal. [20] H. Al-Karawi, J. Al-Emrani, Mohammadand Hedegård, Crack behaviour after high
• The effect of strain hardening, geometry improvement, residual and frequency mechanical impact treatment in welded S355 structural steel, The Inter-
clamping stresses were combined using Goodman mean stress correc- national Conference on Bridge Maintenance, Safety and Management IABMAS (Sub-
mitted and Accepted), Sapporo, Japan, 2020.
tion and Basquin's Equations. Long fatigue lives were obtained which [21] S.J. Maddox, M. Doré, S.D. Smith, A case study of the use of ultrasonic peening for
explained the results from the experiments. The calculated fatigue upgrading a welded steel structure, Welding World 55 (9–10) (2011) 56–67.
lives were found to be in line with the test results. [22] A. Akyel, M. Kolstein, F. Bijlaard, Fatigue strength of repaired welded connections
made of very high strength steels, Eng. Struct. 161 (2018) 28–40.
• The combined effect of TIG-remelting and HFMI-treatment found to [23] A. Akyel, M. Kolstein, F. Bijlaard, Fatigue strength of repaired cracks in base material
be more effective in fatigue life extension than TIG-remelting. How- of high strength steels, J. Constr. Steel Res. 139 (2017) 374–384.
ever, this could not be verified by fatigue testing as no failure at the [24] M. Leitner, Z. Barsoum, F. Schäfers, Crack propagation analysis and rehabilitation by
HFMI of pre-fatigued welded structures, Welding World 60 (3) (2016) 581–592.
weld toe was obtained.
[25] C. Branco, V. Infante, R. Baptista, Fatigue behaviour of welded joints with cracks,
repaired by hammer peening, Fatigue Fract. Eng. Mater. Struct. 27 (9) (2004)
785–798.
[26] V. Infante, C. Branco, A study on the fatigue behaviour of damaged welded joints
Declaration of Competing Interest repaired by hammer peening, ECF13, San Sebastian, 2000, , 2000.
[27] K. Houjou, K. Takahashi, K. Ando, H. Abe, Effect of peening on the fatigue limit of
welded structural steel with surface crack, and rendering the crack harmless, Int.
The authors declare that they have no known competing financial J. Struct. Integr. 5 (4) (2014) 279–289.
interests or personal relationships that could have appeared to influ- [28] R. Fueki, K. Takahashi, Prediction of fatigue limit improvement in needle peened
ence the work reported in this paper. welded joints containing crack-like defects, Int. J. Struct. Integr. 9 (1) (2018) 50–64.
[29] Y. Kudryavtsev, J. Kleiman, A. Lugovskoy, L. Lobanov, V. Knysh, O. Voitenko, G.
Prokopenko, Rehabilitation and repair of welded elements and structures by ultra-
Acknowledgments sonic peening, Welding World 51 (7–8) (2007) 47–53.
[30] H.-P. Günther, U. Kuhlmann, A. Dürr, Rehabilitation of welded joints by ultrasonic
The authors are grateful to the road administration (Trafikverket) impact treatment (UIT), IABSE Symposium Report, vol. 90, International Association
for Bridge and Structural Engineering 2005, pp. 71–77.
and the innovation system agency in Sweden Vinnova (Infra Sweden [31] R. Fueki, K. Takahashi, K. Houjou, Fatigue limit prediction and estimation for the
2030) who financed the whole project. Gratitude is also directed to Nils crack size rendered harmless by peening for welded joint containing a surface
Fredrich from Hamburg University of Technology and for Dr. Mattias crack, Mater. Sci. Appl. 6 (06) (2015) 500.
[32] T. Skriko, M. Ghafouri, T. Björk, Fatigue strength of TIG-dressed ultra-high-strength
Thuvander from Chalmers University of Technology who contributed steel fillet weld joints at high stress ratio, Int. J. Fatigue 94 (2017) 110–120.
generously in fatigue testing, residual stress measurements, and [33] P. Haagensen, et al., Tig dressing of steel weldments for improved fatigue perfor-
microscopy. mance, Offshore Technology Conference, Offshore Technology Conference, 1979.
[34] H. Alkarawi, A. Manai, M. Al-Emrani, Fatigue Life Extension of Welds: A Critical View
on the State of the Art, Chalmers university of technology publications, 2019.
References [35] N. Standard, Material data sheets for structural steel, Norway Technology Center, N-
0306 Oslo, Norway, 2000.
[1] Z. Barsoum, Henry granjon prize competition 2010 winner category C:“design and [36] ESAB, Coreweld C6LF, https://www.esabna.com/us/en/products/filler-metals/
structural integrity” fatigue design of welded structures—some aspects of weld metal-cored-wires-mcaw/mild-steel-wires/coreweld-c6-lf.cfm.
quality and residual stresses, Welding World 55 (11–12) (2011) 2–11. [37] A.M. de Jesus, R. Matos, B.F. Fontoura, C. Rebelo, L.S. da Silva, M. Veljkovic, A compar-
[2] A. Hobbacher, Recommendations for fatigue design of welded joints and compo- ison of the fatigue behavior between S355 and S690 steel grades, J. Constr. Steel Res.
nents. IIW document xiii-2151–07, XV-1254r1–07, Paris 2007 (2007). 79 (2012) 140–150.
12 H. Al-Karawi et al. / Journal of Constructional Steel Research 172 (2020) 106200

[38] E. C. for Standardization, Eurocode 3: Design of steel structures. part 1.9: Fatigue, [41] P. Haagensen, E.S. Statnikov, L. Lopez-Martinez, Introductory fatigue tests on welded
2003. joints in high strength steel and aluminium improved by various methods including
[39] N. Powers, D.M. Frangopol, R. Al-Mahaidi, C. Caprani, Maintenance, safety, risk, man- ultrasonic impact treatment (UIT), IIW Doc 13 (1998) 1748–1798.
agement and life-cycle performance of bridges, Melbourne, Australia, Proceedings of [42] L.L. Martinez, A. Blom, H. Trogen, T. Dahle, Fatigue behaviour of steels with strength
the Ninth International Conference on Bridge Maintenance, Safety and Manage- levels between 350 and 900 MPa influence of post weld treatment under spectrum
ment, 2018. loading, in: A.F. Bloom (Ed.), Proceeding of the North European Engineering and
[40] E.S. Statnikov, V. Muktepavel, A. Blomqvist, Comparison of ultrasonic impact treat- Science Conference,(NESCO), Welded High-Strength Steel Structures, Stockholm,
ment (UIT) and other fatigue life improvement methods, Welding World 46 (3–4) London, EMAS Publishing, 1997.
(2002) 20–32.