ME TECHNICAL PAPER w

x
From Materials Evaluation, Vol. 77, No. 4, pp: 519-528.
Copyright © 2019 The American Society for Nondestructive Testing, Inc.

Experimental Investigation of Weld Root to Throat

Crack Recognition by Double-Probe Ultrasonic Testing
by Ye Zhi*, Ji Bo-Hai†, Wang Jie-Xiu*, and Fu Zhong-Qiu*

ABSTRACT Introduction
Experimental investigations were carried out to Orthotropic steel decks have been widely used for long span
discover a possible approach to detect weld root to bridges because of their low weight, high bearing capacity,
and broad range of application (Xiao et al. 2008; Deng et al.
throat crack in deck-to-rib welded joints by ultra-
2015). Despite having excellent structural properties, different
sonic testing. Twenty-three samples, manufactured discontinuities emerge gradually in the steel deck during its
with artificial cracks to simulate the welded parts service life because of the discontinuities produced in the
of deck-to-rib, were subjected to ultrasonic testing. welding process (Rao et al. 2002), complex local deformation,
increasing traffic volumes, and so on. Fatigue cracking of
Crack features, such as width, length, depth, and
welded joints is one of the most serious problems faced: its
propagation path, were considered. A technique brittle fracture characteristics pose a threat to the safety of
involving double-probe ultrasonic testing was steel bridges. The typical weld root to throat cracks of an
proposed. The transmitted signal’s amplitudes orthotropic steel deck are shown in Figure 1. One initiates at
the weld root and propagates horizontally, and the other may
under different cases were compared. The results
propagate obliquely to the weld throat. These two cracks can
revealed that the ratio we of signal amplitude at only be observed until the fracture of the welded joint occurs,
the crack tip to the signal amplitude of uncracked which may compromise the bridge’s safety and structural
areas was adopted to calculate crack length (a we integrity (Markus et al. 2015). It is therefore necessary to
detect these cracks as early as possible.
value of one-third is suggested). Additionally, a
Nowadays, various nondestructive testing (NDT) tech-
D-A curve comparison technique (where D is the niques are widely accepted for detecting cracks in steel
distance between the probe index to the weld toe bridges, such as magnetic particle inspection (Gang and Chi
and A is the wave signal amplitude) was proposed 2007), eddy current testing (Robert 2012), acoustic emission
testing (Birring 2008), ultrasonic testing (Kenderian 2002;
based on the test results. According to the rela-
Ushakov et al. 2011), and more. Among these techniques, the
tionship between crack depth and signal ultrasonic method has proven to be the most commonly used
amplitude, a formula to calculate crack depth was because ultrasonic waves are sensitive to local hidden cracks
proposed. The difference in curve shapes was used (Kenderian 2002). Different ultrasonic detection techniques
have been studied in a variety of scenarios (Harri et al. 2008;
to distinguish crack propagation paths, which
An et al. 2014; Xie et al. 2016). A crack detection technique
propagated horizontally to the weld and obliquely using the transmission of multisine ultrasonic waves was
to the weld throat. successfully carried out to identify critical crack lengths on a
KEYWORDS: ultrasonic testing, double-probe, weld wind panel (Harri et al. 2008). A wireless ultrasonic wavefield
imaging technique was presented for detecting hidden
root, throat crack, crack features, steel bridge
damage inside a steel box girder bridge (An et al. 2014). The
K-value technique and one-third-length measurement tech-
nique were proposed to detect cracks in a steel box girder
(Xie et al. 2016). These studies show that ultrasonic testing
* College of Civil and Transportation Engineering, Hohai University, has better crack-detection ability. However, as the detection
Nanjing, Jiangsu 210098, P.R. China
† College of Civil and Transportation Engineering, Hohai University, objects become more complex, the limitations of ultrasonic
Nanjing, Jiangsu 210098, P.R. China; email [email protected] testing technology are becoming more obvious, such as

APRIL 2019 • MATERIALS EVALUATION 519

ME TECHNICAL PAPER w
x investigation of weld root to throat cracks

Crack 1

Crack 2

Crack 1

Deck

U-rib

Weld
Crack 2

Figure 1. Weld root to throat cracks in a deck-to-rib welded joint.

insensitivity to microdamage, failure to detect cracks in technique, samples with different crack features such as length,
certain directions, and so on. To overcome the deficiencies in depth, and propagation path were fabricated and detected.
existing ultrasonic detection technology, some advanced UT
techniques have been developed for enriching the capacity of Ultrasonic Transmission Technique
NDT. The nonlinear ultrasonic technique, which uses The ultrasonic transmission technique can be used to judge
nonlinear ultrasonic behavior such as higher-harmonic gener- internal discontinuities in a sample by the amplitude of trans-
ation, subharmonic generation, nonlinear resonance, or mixed mitted ultrasonic signal. The principle of the ultrasonic trans-
frequency response, has been studied as a way of overcoming mission technique to detect weld root cracks of rib-to-deck is
the limitations of lower sensitivity to microdamage in conven- depicted in Figure 2. There is a set of probes for both trans-
tional methods (Jhang 2009). Phased array ultrasonic testing mitting and receiving. The transmitting probe is attached to
was used to detect defective welds because it can improve the upper surface of the deck and the receiving probe is
sensitivity and coverage as well as decrease inspection times attached to the left surface of a U-rib. The ultrasonic wave
(Zhang et al. 2016). To satisfy the requirement for on-line penetrates the interior of the sample and is received by the
detection, lamb wave air-coupled ultrasonic testing has also
been used in some cases (Chakrapani et al. 2012).
Although some types of crack and advanced detection
techniques have been studied in some research, it should be T-wave
noted that weld root to throat cracks in deck-to-rib welded U-rib
joints are often ignored and there is no appropriate technique
Receiver

R-wave
with which to detect these cracks. There are several reasons Third leg
for this: traditional ultrasonic detection techniques have
obvious limitations in detecting the size and position of cracks Transmitter Crack
Second leg
because of the complex local structure of an orthotropic steel
deck. Furthermore, with the high skill level required to
operate such equipment, it is more expensive than conven- Deck β
tional UT, and given the inconvenience of onsite rapid detec-
tion, the technique is unsuitable here. For these reasons, a First leg
double-probe ultrasonic transmission technique was proposed
to detect weld root to throat cracks inside rib-to-deck connec- Figure 2. The detection principle of the ultrasonic transmission
tions. To validate the feasibility of the proposed detection technique.

520 MATERIALS EVALUATION • APRIL 2019

receiving probe, and then the transmitted ultrasonic signal The range of angles of incidence a, between 27.6° to 57.7°,
peak amplitude can be observed on the oscilloscope. The can be obtained:
transmission ultrasonic wave signal of the receiving probe will cL 1 2730
be reduced or disappeared when part of the transmission α I = arcsin = arcsin = 27.6°
cL 2 5900
ultrasonic wave is stopped by the crack generated at the weld (1)
c 2730
root. Therefore, the technique can be used to detect weld root α II = arcsin L1 = arcsin = 57.7°
cracks effectively. cS 2 3230
In this study, an angle probe was chosen to detect weld
root cracks of rib-to-deck joints. Considering that the According to Equation 2, the refraction angle b can be
accuracy of the ultrasonic inspection could be affected by the obtained and the range is 33.2° to 90°. The calculated K-value
characteristics of the angle probe, parameters such as the is greater than 0.66 based on Figure 3.
K-value, frequency, and chip size need to be of an appropriate sin α L sin βS
magnitude. The K-value is defined as the tangent value of (2) =
cL1 cS 2
angle b between the incident wave and an interface normal
line (Figure 3). To ensure that the scanning range of the ultrasonic beam
When ultrasonic waves propagate through the interface of could cover most of the weld and both the minimum distance
two different media, two kinds of waves (longitudinal and of A (probe index to weld toe) and B (probe receiving point
shear) are transmitted to the measuring object, and shear to weld toe) are longer than the probe frontier length, the
waves are used in the measurement of crack characteristics K-value range was chosen to be within 0.7 to 0.9. A K-value
(Vladišauskas et al. 2005; Carodiskey and Meyer 1997). To of 0.8 for the transmitting probe was selected in this study.
obtain only the shear wave in the second medium, the angle According to the law of reflection, the angle between the
of incidence a should be set as an angle between the first and beam axis of the receiving probe and interface normal is
second critical angle. In this test, the wedge is made of acrylic 63° and the corresponding K-value is 1.96, so the K value was
glass. The velocity of a longitudinal wave in the acrylic glass selected as 2.0 here.
cL1 is 2730 m/s. Furthermore, the detected sample is steel, The frequencies used by the probes are between 1.25
and the velocity of ultrasonic wave in steel is about cL2 = and 20 MHz. In this test, a 5 MHz probe was used and a
5900 m/s for longitudinal wave and cS2 = 3230 m/s for 4 mm × 4 mm chip size was chosen. The properties of the
shear wave. In this case, the first and second critical probe were calibrated on a standard test block before use
angle can be deduced (Vladišauskas et al. 2005). (Table 1).

Experimental Schemes
Probe

L Samples
Three types of samples (cracked standard sample, uncracked
α K = tan β = LH contrastive sample, and sample for validation purposes) were
fabricated for the experiment (Table 2). All samples were
made from Q345 steel, in accordance with Chinese Specifica-
H β S tion GB/T 700 (People’s Republic of China 2006). The
dimensions of the deck segment were 150 mm × 60 mm
(120 mm), the thickness of the deck plate was 12 mm, and
the thickness of the rib wall was 6 mm. The sample sizes are
L
shown in Figure 4a. During welding of the sample, carbon
dioxide-protected welding was used. Considering the depth of
Figure 3. K-value calculation diagram. the unmelted area was inconsistent because of the actual

TABLE 1
Probe parameters
Function Frequency Size of chip Calculated Real Length of probe
(MHz) (mm) K-value K-value frontier (mm)
Pitch 5 4×4 0.8 0.78 4.55
Catch 5 4×4 2.0 2.20 5.50

APRIL 2019 • MATERIALS EVALUATION 521

ME TECHNICAL PAPER w
x investigation of weld root to throat cracks

TABLE 2
Crack parameters
Type of sample Crack Type Width Length Depth Deck width
number (mm) (mm) (mm) (mm)
Cracked standard sample A1-L1 Crack 1 0.20 20 Full cracking to weld toe 60
Cracked standard sample A1-L2 Crack 1 0.20 35 Full cracking to weld toe 120
Cracked standard sample A1-L3 Crack 1 0.20 50 Full cracking to weld toe 120
Cracked standard sample A1-D1 Crack 1 0.20 60 2 60
Cracked standard sample A1-D2 Crack 1 0.20 60 4 60
Cracked standard sample A1-D3 Crack 1 0.20 60 6 60
Cracked standard sample A1-W1 Crack 1 0.10 60 4 60
Cracked standard sample A1-W2 Crack 1 0.15 60 4 60
Cracked standard sample A1-W3 Crack 1 0.20 60 4 60
Cracked standard sample A2-L1 Crack 2 1.00 20 Full cracking to weld throat 60
Cracked standard sample A2-L2 Crack 2 1.00 35 Full cracking to weld throat 120
Cracked standard sample A2-L3 Crack 2 1.00 50 Full cracking to weld throat 120
Cracked standard sample A2-D1 Crack 2 0.20 60 1.5 60
Cracked standard sample A2-D2 Crack 2 0.20 60 3.0 60
Cracked standard sample A2-D3 Crack 2 0.20 60 4.5 60
Cracked standard sample A2-W1 Crack 2 0.10 60 3 60
Cracked standard sample A2-W2 Crack 2 0.15 60 3 60
Cracked standard sample A2-W3 Crack 2 0.20 60 3 60
Uncracked contrastive sample B1 – – – – 60
Uncracked contrastive sample B2 – – – – 60
Uncracked contrastive sample B3 – – – – 120
Sample for validation purposes C1 Crack 1 1.00 20/50 Full cracking to weld toe 120
Sample for validation purposes C2 Crack 2 1.00 20/50 Full cracking to weld throat 120

U-rib U-rib
6 mm
Crack 1 Crack 2

U-rib
10 0 m m

7 mm Crack 2
Deck Unmelted zone Deck Unmelted zone
Z
Crack 1 X Y
8 mm
No crack
103°
Weld 12 mm

Deck m
50 m
60/120 mm

m
10 0 m
Unmelted zone
(a) (b)

Figure 4. Sample details: (a) dimensions of a cracked sample; (b) two categories of artificial cracks.

522 MATERIALS EVALUATION • APRIL 2019

 X Z Z X Z

Y Y
Deck
U-rib

Receiver
U-rib U-rib Scan direction
Receiver Crack Crack
Crack
Receiver
Weld
Weld Weld
Crack tip

Scan direction
Transmitter

Transmitter
Transmitter

A1 A2 A3
B1 B2 B3 B4
Ruler

Ruler
Deck Deck Uncracked area
(a) (b)

Figure 5. The scan direction of the probe: (a) before and after scanning; (b) left and right scanning.

fabrication error, the weld root of all samples (both the measurement point was measured three times and nine D-A
standard and contrast samples) were grooved to a depth of curves established. Following these, the nine curves for each
2.00 mm to nullify the impact of the unmelted area. Two sample could be used to fit a D-A curve of the sample through
types of cracks (crack 1 and 2) were fabricated considering nonlinear bi-gaussian function fitting. The measurement tech-
the width, depth, length, and propagating path, and the main nique used to derive the D-A curve in cracked samples with
cracks in the standard samples were manufactured by elec- different widths and depths is the same as that used on
trical discharge machining (EDM), except that samples A2- uncracked samples, and the D-A curve of an uncracked
L1, A2-L2, and A2-L3 were cut by machine. Considering the sample was defined as curve R (reference curve) and the
characteristics of machine cutting, the crack depth of samples cracked sample as curve F (feature curve). By comparing
A2-L1, A2-L2, and A2-L3 were larger than in other samples. curves R and F, crack features such as width, depth, and incli-
Furthermore, regarding the direction of crack 2 being cut nation angle could be judged.
toward the weld throat of these samples, with a smaller poten- The technique used to measure crack length is shown in
tial propagating length than others, the depth of fabricated Figure 5b. Firstly, four measurement points (B1, B2, B3, B4)
cracks in crack 2 is shorter than crack 1, as shown in Table 2. located on the left-hand side of the uncracked area, left-hand
side of the crack tip, right-hand side of the uncracked area,
Test Procedure and right-hand side of the crack tip, were selected and
The measurement of uncracked contrastive samples was marked. Then the transmitting and receiving probe was
conducted first (Figure 5a). Three measurement points moved perpendicularly to the weld repeatedly and independ-
(A1, A2, A3), equidistantly spaced along the weld of the ently until the peak amplitude of the wave signal could be
sample, were selected and marked. The transmitting and received by the two probes. Thereafter, the location of the
receiving probes were repeatedly moved in a direction receiving probe was fixed. After that, the two probes were
perpendicular to the weld independently until the peak ampli- moved synchronously from the uncracked area to the crack
tude of wave signal was obtained, then the location of the tip along a path parallel to the weld. During this movement,
receiving probe was fixed. Then the transmitting probe was the two probes were kept the same distance apart. The corre-
moved toward (and away from) the weld joint in 1 mm steps, sponding signal amplitudes at each of the four measurement
and the corresponding wave signal amplitude was recorded at points were recorded, and the relationship of the ratio we of
each step. Following the recording step, the D-A curve of the signal amplitude of crack tip to the signal amplitude of the
measurement point can be established by the distance uncracked area could be established, to allow calculation of
between the probe index to the weld toe (D) and the wave the crack length from the value of we.
signal amplitude (A). To ensure accuracy and reliability, each

APRIL 2019 • MATERIALS EVALUATION 523

ME TECHNICAL PAPER w
x investigation of weld root to throat cracks

Crack Width
100 Considering that the actual crack width will change during its
B1-A1 B2-A3 propagation, the influence of crack width on the results of the
B1-A2 B3-A1
80 B1-A3 B3-A2 inspection technique should be investigated first. Samples
B2-A1 Mean A1-W1, A1-W2, A1-W3, A2-W1, A2-W2, and A2-W3 were
Mean ±2 × STD
Reference curve selected to measure crack width. The D-A feature curves
Amplitude (%)

60
considering different crack widths are shown in Figure 7:
for the same crack, the D-A feature curves at different crack
40 widths are similar. The location of the probe index correspon-
ding to the signal peak amplitude on these feature curves is
20 also similar. These indicate that crack width has little effect on
the inspection technique. Therefore, the influence of crack
0 width is neglected in this test.

4 6 8 10 12 14 16 18 20 22 24 26 Crack Length
Distance from probe index to weld toe (mm) The crack length is the distance between the two crack tips.
To measure the crack length, samples A1-D1, A1-D2, A1-D3,
Figure 6. D-A reference curves of uncracked samples. A2-D1, A2-D2, and A2-D3 were selected and detected in this
test. The maximum signal amplitude was set approximately
90% high by adjusting the sensitivity. The probe was moved
Analysis Results horizontally along the cracks, and the changes in signal ampli-
tude recorded. Figure 8 shows the results from different meas-
D-A Reference Curve of the Uncracked Sample
urement points.
The gaussian function was used to fit the test data of the As shown in Figure 8, when the probe is on the uncracked
uncracked sample. In the reference curve fitting process, the area, the signal amplitude is the highest, and when the probe
data from the B2-A2 and B3-A3 measurement points were not is moved toward the crack tip, the signal amplitude decreased
adopted, because defects other than the prefabricated cracks rapidly. The peak signal amplitude is approximately 3× the
were found in the weld, which resulted in major errors in the minimum value. The proportional relationship can be
test data. The fitted D-A reference curve (curve R) of an expressed as a ratio using we. The crack tip can be identified
uncracked sample is shown in Figure 6 and the goodness of fit according to the value of we and Equation 3, and then the
R2 = 0.94. This reference curve can be used for comparison crack length can be measured.
with the feature curve (curve F).

100 100
Crack 1 Uncracked
Crack 2 Uncracked
W = 0.10 mm W = 0.10 mm
80 W = 0.15 mm 80 W = 0.15 mm
W = 0.20 mm W = 0.20 mm
Amplitude (%)

Amplitude (%)

60 60

40 40

20 20

0 0
4 8 12 16 20 24 28 4 8 12 16 20 24 28
(a) Distance from probe index to weld toe (mm) (b) Distance from probe index to weld toe (mm)

Figure 7. D-A feature curve for various crack widths (W = crack width): (a) crack 1; (b) crack 2.

524 MATERIALS EVALUATION • APRIL 2019

 100 100
L = 20 mm L = 35 mm L = 50 mm L = 20 mm L = 35 mm L = 50 mm

80 80
Crack Crack tip Crack Crack tip
Crack tip Crack tip
Amplitude (%)

Amplitude (%)
Right Right
B4 B4
60 B3 60 B3
B2 B2
B1 Uncracked area B1 Uncracked area
Left Left
40 40

20 20

0 0
B1 B2 B3 B4 B1 B2 B3 B4
(a) The location of probe index (b) The location of probe index

Figure 8. Wave signal amplitude at measurement points (L= crack length): (a) crack 1; (b) crack 2.

(3) R rtip = ωe Rmax

r
depths. Curves F and D for different crack depths are shown
in Figure 9.
where Figure 9 shows that the feature curves for different crack
Rrtip is the signal amplitude of crack tip in the real test, and depths varied: compared with curve R, curve F was right-
Rrmax is the peak signal amplitude of the uncracked area in shifted, and this increased with increasing crack depth. The
the real test. cause of this was related to the crack propagation path. The
two types of cracks were both initiated at the weld root and
Crack Depth propagated horizontally or obliquely to the weld. When the
To detect the crack depth and type, a curve D (difference transmitting probe moved toward and away from the weld,
curve) was proposed by subtracting curve R (reference curve) the wave was first stopped by the crack, then passed through
from curve F (feature curve). Samples A1-D1, A1-D2, A1-D3, the penetrable range of the weld area. As the crack depth
A2-D1, A2-D2, and A2-D3 were selected to measure crack increased, the time that wave entered the penetrable area

120 120
Crack 1 Crack 2

80 80
Amplitude (%)

Amplitude (%)

Deviation Overlap
40 Curve R 40 Curve R

0 0

Crack depth Curve F Curve D Crack depth Curve F Curve D

–40 –40
Extreme point 2 mm Extreme point 1.5 mm
4 mm 3.0 mm
6 mm 4.5 mm
–80 –80
4 8 12 16 20 24 28 32 4 8 12 16 20 24 28 32
(a) Distance from probe index to weld toe (mm) (b) Distance from probe index to weld toe (mm)

Figure 9. D-A feature curves for different crack depths: (a) crack 1; (b) crack 2.

APRIL 2019 • MATERIALS EVALUATION 525

ME TECHNICAL PAPER w
x investigation of weld root to throat cracks

Distance from probe index to weld toe (mm)

18 0
Crack 1 Crack 1
Crack 2 Crack 2
16

Difference of amplitude (%)

–20
14 y = –10.39–7.35z

x = 9.94+0.50z
12 –40

10 x = 9.36+0.51z
–60
8
y = –16.74–7.54z

6 –80
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
(a) Crack depth (mm) (b) Crack depth (mm)

Figure 10. Formulae for crack depth: (a) crack depth versus distance from probe index to weld toe; (b) crack depth versus difference of
amplitude.

increased. The distance from the probe index to the weld toe the probe index to the weld toe and y is the amplitude. The
corresponding to the highest amplitude also increased. It is linear relationship between crack depth z and (x,y) was fitted,
seen that the right-hand segment of curve F deviated from and the formulae that were used to calculate crack depth were
curve R for crack 1, and overlapped curve R for crack 2. established (Figure 10). In actual testing, the x- and y-values of
According to the particular relationship between curves R and the extreme point were substituted into the corresponding
F, cracks 1 and 2 could be distinguished; however, a short- formula to obtain two crack depths. The calculated crack depth
coming existed in this technique in that, when the crack depth was taken as the mean of the two calculated depths.
was 2 mm or less, the two types of cracks cannot be distin-
guished precisely. This was deemed to have been due to a Methodological Verification
limit to the resolution of the test equipment.
Curve D, for different crack depths, has an extreme point, Judgment of Crack Type
as shown in Figure 9. The relationship between the extreme To verify the correctness of the judgment process mentioned
point and crack depth was investigated. Each extreme point previously for crack features and applicability of the ultrasonic
corresponded to paired (x,y) data, where x is the distance from transmission technique, two samples (C1and C2) were

150 mm 150 mm
162°

Crack 1 Crack 2
28°
U-rib U-rib
120 mm

120 mm
50 mm

50 mm
20 mm

20 mm

Deck 8 mm Deck
162°

28°

(a) (b)

Figure 11. Dimensions of verification samples: (a) sample C1; (b) sample C2.

526 MATERIALS EVALUATION • APRIL 2019

selected. Differing from the above test sample, the edge shape to T1, which deviated from the position of the real crack tip.
of the crack propagation in samples C1and C2 was trape- The tests were manual in nature and thus prone to error. As
zoidal in order to simulate the real crack shape. Figure 11 the crack length increased, the relative error decreased.
shows the dimensions of samples C1 and C2.
Three measurement points (T1, T2, T3) were selected Validation of Crack Depth
equidistantly on the edge of the crack (Figure 12). Then these The x- and y-values of the extreme point were substituted into
points were detected, one-by-one, using the crack depth tech- the aforementioned formula in Figure 10. Two calculated
nique previously mentioned, and the D-A feature curve and values of crack depth, z1 and z2, were obtained by the
D-A difference curve at each measurement point were formulas in Figure 10a and Figure 10b, respectively. The
obtained separately (Figure 13). mean value of z1 and z2 was compared with the actual depth
It can be seen from Figure 13a that the right-hand to judge the applicability and accuracy of the formula.
segment of curve F deviated from curve R, which could be The experimental results are listed in Table 3.
used to judge the crack type 1. And the right-hand segment of The results in Table 3 indicate that the calculated crack
curve F overlapped curve R, which could verify the presence depth was accurate to within a relative error of 5%. The
of crack type 2. The crack type judged by the difference error arose because the ultrasonic beam underwent a
between curves F and D was consistent with the actual crack certain diffusion angle, and the transmission wave thus
type; thus, the technique was deemed able to distinguish deviated from the actual crack tip location. Meanwhile,
different kinds of weld-root cracks in most cases. some defects such as slag and stoma in welds and lamina-
In the validating test, the value of we was used to identify tions in rib or deck steel materials may have affected the
the crack tip. The measured position of the crack tip was close accuracy of measurement.

Crack tip (mm) Crack tip (mm)

3 3 3 3 3 3 3 3

Crack depth (mm)

Crack depth (mm)

2.4 1.9
3.2
4.0 4.4
5.6

T1 T2 T3 T1 T2 T3
(a) (b)

Figure 12. Measurement point and crack tip locations: (a) crack 1; (b) crack 2.

100 100

80 80

60 60
Curve R Curve R
Amplitude (%)

Amplitude (%)

40 40
Deviation Overlap
20 20

0 0

–20 Crack depth Curve F Curve D

–20 Crack depth Curve F Curve D
T1 2 mm T1 1.9 mm
–40 T2 4 mm –40 T2 3.2 mm
Extreme point Extreme point
T3 5.6 mm T3 4.4 mm
–60 –60
4 8 12 16 20 24 28 32 4 8 12 16 20 24 28 32
(a) Distance from probe index to weld toe (mm) (b) Distance from probe index to weld toe (mm)

Figure 13. Validation test data: (a) crack 1; (b) crack 2.

APRIL 2019 • MATERIALS EVALUATION 527

ME TECHNICAL PAPER w
x investigation of weld root to throat cracks

TABLE 3
Comparison of crack depths
Crack Distance Difference z1 z2 Average Actual Error
number from probe in amplitude (mm) (mm) depth depth (%)
index to toe (%) (mm) (mm)
(mm)
C1-T1 10.6 –28.8 2.4 2.5 2.45 2.4 2
C1-T2 11.8 –36.7 4.8 3.6 4.2 4 5
C1-T3 12.4 –49.5 5.9 5.3 5.6 5.6 0
C2-T1 10.9 –31.2 1.9 1.9 1.9 1.9 0
C2-T2 11.5 –45.3 3.1 3.8 3.5 3.2 9
C2-T3 12.3 –47.8 4.7 4.1 4.4 4.4 0

Conclusions Deng, Y., Y. Liu, D. Feng, and A.Q. Li, 2015, “Investigation of Fatigue
Performance of Welded Details in Long-Span Steel Bridges using Long-
Several techniques for detecting crack features in steel bridge Term Monitoring Strain Data,” Structural Control and Health Monitoring,
decks were discussed. The experimental results have shown Vol. 22, No. 11, pp. 1343–1358.
that the double-probe ultrasonic transmission technique can Erf, Robert (ed.), 1974, Holographic Nondestructive Testing, Elsevier
(Academic Press), New York, NY.
detect the width, length, depth, and type of fatigue crack. The
Gang, T., and D.Z. Chi., 2007, “Novel Approach to Enhancement of Ultra-
following conclusions may be drawn from this study: sonic TOFD B-Scan Image for Measurement of Weld Crack,” Science and
1. The feature curves for different crack widths were consis- Technology of Welding & Joining, Vol. 12, No. 1, pp. 87–93.
tent, and the influence of crack width on the test can be Harri, Kristof, P. Guillaume, and S. Vanlanduit, 2008, “On-line Damage
neglected when using the chosen ultrasonic transmission Detection on a Wing Panel Using Transmission of Multisine Ultrasonic
Waves,” NDT & E International, Vol. 41, No. 4, pp. 312–317.
method.
Jhang, Kyung-Young, 2009, “Nonlinear Ultrasonic Techniques for Nonde-
2. The amplitude ratio we of crack tip to uncracked zone can structive Assessment of Micro Damage in Material: A Review,” Interna-
be used to locate a crack tip as verified by experimental tional Journal of Precision Engineering & Manufacturing, Vol. 10, No. 1,
results (the chosen value of we was one-third). pp. 123–135.
3. A formula to calculate crack depth was proposed on the Kenderian, S., 2002, “Advanced Ultrasonic Techniques to Determine the
Structural Integrity of Rail Steel,” Ph.D. thesis, The John Hopkins Univer-
basis of the relationship between the extreme point and sity, Baltimore, MD.
crack depth. The results of verification testing showed that People’s Republic of China, 2006, GB/T 700-2006, Carbon Structural
crack depth can be detected to within a relative error of 5%. Steels, National Standard of the People’s Republic of China, General
4. Crack type can be determined according to the relation- Administration of Quality Supervision, Inspection and Quarantine of the
People’s Republic of China.
ship between the feature curve and reference curve. Based
Rao, B.P.C, B. Raj, T. Jayakumar, and P. Kalyanasundaram, 2002, “An Arti-
thereon, if the right-hand segment of the feature curve ficial Neural Network for Eddy Current Testing of Austenitic Stainless
deviates from the reference curve, the crack initiates at the Steel Welds,” NDT & E International, Vol. 35, No. 6, pp. 393–398.
weld root and propagates horizontally; if the right-hand Ushakov, V.M., D.M. Davydov, and L.I. Domozhirov, 2011, “Detection
segment of the feature curve is overlapped with the refer- and Measurement of Surface Cracks by the Ultrasonic Method for Evalu-
ating Fatigue Failure of Metals,” Russian Journal of Nondestructive Testing,
ence curve, the crack propagates obliquely to the weld Vol. 47, No. 9, pp. 631–641.
throat. The results of the verification test showed that the Vladišauskas , A., L. Jakevičius, and J. Butkus, 2005, “The Use of Ultra-
technique was effective. sound for Determination of the Thickness of Glass Panes of Windows”
Ultragarsas, Vol. 10, No. 7, pp.1291–1296.
ACKNOWLEDGMENTS Wohlschlögel, Markus, Gunter Gläßel, Daniela Sanchez, Andreas Schüßler,
The authors wish to appreciatively acknowledge support provided through Alexander Dillenz, David Saal, and Peter Mayr, 2015, “Characterization of
the National Natural Science Fund No. 51678216 and No. 51478163. Nitinol Laser-Weld Joints by Nondestructive Testing,” Journal of Materials
Engineering and Performance, Vol. 24, No. 12, pp. 4991–4996.
REFERENCES Xiao, Z.G., K. Yamada, S. Ya, and X.-L. Zhao, 2008, “Stress Analyses and
Fatigue Evaluation of Rib-to-Deck Joints in Steel Orthotropic Decks,”
An, Yun-Kyu, H. Song, and H. Sohn, 2014, “Wireless Ultrasonic Wavefield
International Journal of Fatigue, Vol. 30, No. 8, pp. 1387–1397.
Imaging via Laser for Hidden Damage Detection inside a Steel Box Girder
Bridge,” Smart Materials & Structures, Vol. 23, No. 9, pp. 639–650. Xie, Fa-Xiang, B.-H. Ji, Z.-Y. Yuanzhou, Z.-Q. Fu, and H.-B. Ge, 2016,
“Ultrasonic Detecting Method and Repair Technology Based on Fatigue
Birring, Anmol S., 2008, “Ultrasonic Phased Arrays for Weld Testing,”
Crack Features in Steel Box Girder,” Journal of Performance of Constructed
Materials Evaluation, Vol. 66, No. 3, pp. 282–284.
Facilities, Vol. 30, No. 2; doi: 10.1061/(ASCE)CF.1943-5509.0000725.
Carodiskey, T.J, and P.A. Meyer, 1997, “Thickness Measurement in Mate-
Zhang, Haibing, J. Sun, and H. Sun, 2016, “Experimental Research Based
rials of Unknown Acoustic Velocity,” NDT.net, Vol. 2, No. 10, pp. 1–3.
on Phased Array Ultrasonic Inspection Technology for Weld Joint Inspec-
Chakrapani, Sunil Kishore, M. Janardhan Padiyar, and K. Balasubrama- tion of a Type of Aircraft Standing Bar,” Aeronautical Manufacturing Tech-
niam, 2012, “Crack Detection in Full Size Cz-Silicon Wafers Using Lamb nology, Vol. 59, No. 22, pp. 89–91.
Wave Air Coupled Ultrasonic Testing (LAC-UT),” Journal of Nondestruc-
tive Evaluation, Vol. 31, No. 1, pp. 46–55.

528 MATERIALS EVALUATION • APRIL 2019