Professional Documents
Culture Documents
NDT&E International
journal homepage: www.elsevier.com/locate/ndteint
Rail Research Centre, The University of Birmingham Edgbaston, Birmingham B15 2TT, UK
TSC Inspection Systems, 6 Mill Square, Featherstone Road, Milton Keynes MK12 5RB, UK
a r t i c l e in f o
a b s t r a c t
Article history:
Received 2 August 2008
Received in revised form
13 December 2008
Accepted 18 December 2008
Available online 20 February 2009
High-speed ACFM tests were carried out using a rotary test piece that contained spark-eroded notches.
The ACFM sensor detected the induced notches during inspection at 121.5 km/h. The recorded signal
remained unaffected by the increases in inspection speed under constant lift-off. To simulate actual rail
inspection conditions at high speed, further tests were carried out using a spinning rail rig and a set of
rails that contained spark-eroded notches of various shapes and sizes up to a speed of 32 km/h.
Although, the ACFM sensor successfully detected the majority of the notches, the signal obtained was
affected by lift-off variations.
& 2009 Elsevier Ltd. All rights reserved.
Keywords:
Rail inspection
ACFM
Crack detection
Rolling contact fatigue
Electromagnetic
1. Introduction
Ultrasonic rail inspection is carried out by a variety of different
instruments ranging from hand-held devices, through dualpurpose road/track vehicles to test xtures that are towed or
carried by dedicated rail cars. Unfortunately, the performance of
existing conventional ultrasonic probes in detecting small
(o4 mm) surface defects such as head checks and gauge corner
cracking is inadequate during high-speed inspection, particularly
above 15 km/h. Fig. 1 shows a rail containing typical rolling
contact fatigue (RCF) cracking. Furthermore, the presence of larger
and more critical internal defects can be shadowed by clusters of
smaller surface cracks during ultrasonic testing. Inspection
systems based on the simultaneous use of conventional ultrasonic
transducers with magnetic ux leakage (MFL) sensors have a
higher probability of detecting smaller near-surface and surfacebreaking defects in the rail head. However, as inspection speed
increases, the performance of MFL sensors tends to deteriorate
rapidly due to a reduction in the magnetic ux density arising
from the eddy current excitation [13]. The use of higher
magnetic elds to compensate for the reduction in magnetic ux
density may not be straight forward due to the potential
interference with the electronic equipment installed on rail
tracks. More recently, pulsed eddy current (PEC) probes have
been added on certain ultrasonic test trains to offer increased
sensitivity in the detection of surface defects at inspection speeds
Corresponding author. Tel.: +44 121 414626; fax: +44 121 414429.
ARTICLE IN PRESS
M.Ph. Papaelias et al. / NDT&E International 42 (2009) 328335
329
Bz
Clockwise flow
gives Bz peak
Uniform input current
z
x
Anti-clockwise
flow gives Bz
trough
Fig. 1. Typical RCF cracking found in rails (photograph is courtesy of Dr. Garnham).
electromagnetic properties of the metal under inspection. Typically, the maximum skin depth achieved during inspection with
ACFM systems varies from 0.1 mm for carbon steels to 6 mm for
stainless steel.
Current lines
close together
gives Bx peak
2. ACFM principle
Bx
The ACFM technique, also known as the electric current
perturbation (ECM) method, is based on the principle that an
alternating current (AC) can be induced to ow in a thin skin near
the surface of any conductor. By introducing a remote uniform
current into an area of the component under test, when there are
no defects present the electrical current will be undisturbed. If a
crack is present, the uniform current is disturbed and the current
ows around the ends and down the faces of the crack. Because
the current is an alternating current it ows in a thin skin close to
the surface and is unaffected by the overall geometry of the
component.
Associated with the current owing in the surface is a
magnetic eld above the surface which, like the current in the
surface, will be disturbed in the presence of a defect. An important
factor of the ACFM technique is its ability to relate measurements
of the magnetic eld disturbance to the size of defect that caused
that disturbance. The development of the method came from a
combination of research studies, which provided mathematical
modelling of the magnetic eld rather than electrical elds, and
advances in electronics and sensing technology [1119]. Despite
the fact that the magnetic eld above the surface is a complex
three-dimensional eld, it is possible, by choosing suitable
orthogonal axes, to measure components of the eld that are
indicative of the nature of the disturbance and which can be
related to the physical properties of any cracks present. Fig. 2
presents a plan view of a surface-breaking crack where a uniform
AC current is owing. The eld component denoted by Bz
corresponds to the poles generated as the current ows around
the ends of the crack introducing current rotations in the plane of
the component. These responses are principally at the crack ends
and are indicative of a crack length. The eld component denoted
by Bx corresponds to the reduction in current surface density as
the current ows down the crack and is indicative of the depth of
the defects. Generally, the current is introduced perpendicular to
the expected direction of cracking. In practice, special probes have
been developed which contain a remote eld induction system,
for introducing the eld into the component, together with special
combined magnetic eld sensors that allow accurate measurement of the components of the magnetic eld at the same point in
space. The probe requires no electrical contact with the component and can therefore be applied without the removal of surface
contaminants.
ACFM probes are available as standard pencil probes and
multi-element array probes. These probes can be customised to
optimise inspection of particular structural components and
maximise the probability of detection (PoD) of critical-sized
defects. ACFM pencil probes can detect surface-breaking defects
in any orientation. Nonetheless, in order to size defects, they need
to lie between 01 and 301 to the direction of travel of the probe.
This drawback is overcome in ACFM arrays by incorporating
various eld inducers to allow a eld to be introduced within the
inspected surface in other orientations. This is particularly useful
in situations where the crack orientation is unknown or variable.
In this case, additional sensors, denoted by By, are also
incorporated in order to take full advantage of the additional
input eld directions.
ARTICLE IN PRESS
330
2mm deep
Rotary sample
ACFM Probe
Fig. 4. Experimental setup for the turning lathe tests using the rotary test piece.
data in real time. The data logged during these tests were the
ACFM probe signal (i.e. analogue-to-digital conversion of probe
voltage) with time. Data were collected for various speeds,
starting from equivalent surface speeds of 4.5 km/h (at 100 rpm)
up to 121.5 km/h (at 3000 rpm) as discussed furthur.
4mm deep
4mm deep
2mm deep
Fig. 3. Schematic diagram of the rotary test piece with four spark-eroded notches.
ARTICLE IN PRESS
M.Ph. Papaelias et al. / NDT&E International 42 (2009) 328335
890
81km/h
880
A/D Conversion
870
860
850
840
830
2mm notch
880
A/D Conversion
870
860
850
840
830
820
810
0.00
0.01
0.03
0.02
0.04
0.05
Time / s
Fig. 7. ACFM data plots showing the effect of increasing lift-off at 81 km/h.
890
880
870
860
850
840
830 4mm notch
820
810
800
790
780
770
760
750
4mm notch
740
730
720
710
700
690
0.00
0.01
2mm
5mm
2mm
notch
0.02
0.03
0.04
0.05
Time / s
820
4mm notch
810
0.00
0.01
0.02
0.03
Time / s
0.04
0.05
20.25km/h
890
121.5km/h
880
A/D Conversion
2mm
5mm
890
A/D Conversion
331
870
860
850
840
830
820
810
0.00
0.01
0.02
0.03
0.04
0.05
Time / s
Fig. 6. ACFM data plots showing the resulting signals at 20.25 and 121.5 km/h with
0.8 mm lift-off.
ARTICLE IN PRESS
332
the rig as seen from the plot in Fig. 12. The fact that lift-off
variations existed within each test rail due to bowing, affected the
ACFM measurements, since the trolley used was designed to
compensate for lift-off variations from rail-to-rail and not within
each rail. Lift-off occurring due to bowing in the individual test
rails meant that the probe stand-off was highest at the centre of
each rail where the notches were present. This sort of lift-off
variation is not expected in real-life inspection of rails.
Fig. 11. Experimental setup during the spinning rail rig tests. The ACFM probe can
be seen in the foreground.
Fig. 10. Schematic showing the test rails for the rail rig tests. Vd vertical depth of the notch in mm, ha horizontal angle of orientation in degrees, va vertical angle of
orientation in degrees. The values in mm shown on top and at the ends of each rails illustrate the probe lift-off at that particular section during the rig tests.
ARTICLE IN PRESS
M.Ph. Papaelias et al. / NDT&E International 42 (2009) 328335
that the system could detect every notch induced on the test rails.
Following these tests, the probe was installed on the customised
trolley and the rig was rotated at 1.6 km/h, progressively
increasing the speed of the spinning rig up to 32 km/h.
The probe was positioned in such a way that the induced
current owed in the longitudinal direction of the rail head. As
shown in Fig. 10, the notches to be inspected are at a 01 or 351
surface angle with respect to the probe orientation and 901 or 251
to the railhead surface. Although the main objective of these tests
was to verify the capability of detection of the ACFM probe at high
speeds under simulated rail inspection conditions, for these
orientations, the ACFM system used should also be able to offer
a quantitative measure of the depth of the notches detected.
However, due to lift-off variations that occurred during testing it
was impossible to obtain correct rankings for all inspected
notches, while certain notches were not positively detected.
The plots in Fig. 13 show the data obtained for rails #5 and #3
at a speed of 2.3 km/h. The effect of lift-off variation can be clearly
Probe Lift-off / mm
7
6
5
4
5
6
7
6
7
0
0
200
400
600
800
1000
1200
Distance / cm
Fig. 12. Probe lift-off variation for a complete rail rig rotation. The letters signify
the lift-off for each test rail.
333
seen by the changes in the overall ACFM signal plot. However, the
induced notches were clearly detected in both rails. Rail #5
contained three groups of clustered notches (notches with depths
2, 10 and 4 mm) as demonstrated in Fig. 10. Each clustered group
of notches had a different spacing between the notches of the
cluster of 5, 10 and 15 mm, respectively. The reason for using
clustered notches was to assess the capability of the sensor in
resolving the presence of individual defects when clustered. As
seen in the ACFM data plot in Fig. 13, despite the fact that all three
clusters of induced notches have been detected, it has not been
possible to resolve individual notches within the clusters due to
the high probe lift-off involved during the experiments (more
than 4 mm). The plot in Fig. 14 shows the ACFM data acquired for
rails #5 and #3 at an inspection speed of 32 km/h. The data
acquired at an inspection speed of 2.3 km/h is virtually identical to
that obtained at 32 km/h as expected from the lathe tests.
Evaluation of the ACFM data acquired for the pocket-shaped
articial notches in rails #7 and #8, which were designed to
resemble more closely actual RCF cracking, revealed that the 2 and
4 mm deep pocket-shaped notches were not detected at all while
the deeper and longer pocket-shaped notches (i.e. 6, 10 and 15 mm
deep) produced a very weak signal and were barely detectable. It
should be noted here that the notches in rails #1#6 had a surface
length that run from the edge to the centre of the railhead, while
the pocket-shaped notches had a signicantly smaller surface
length (i.e. 10 mm for the 2 mm deep notch). The data plot in
Fig. 15 shows the ACFM signal obtained for the three pocketshaped notches in rail 8. The signal obtained is much weaker than
that obtained for the notches in rail #3, despite the fact that the
lift-off variation is similar for both rails (45 mm). This
discrepancy, when comparing the intensity of the signals for
the pocket-shaped notches with the at-proled notches in the
other rails at similar lift-off has been attributed to the smaller
surface length of the pocket-shaped notches as well as the lift-off
effect. In ACFM measurements, the reduction in the signal with
increasing lift-off is greater as the surface length of the defect
decreases.
Fig. 13. Raw and normalised ltered data plots for test rails #5 and #3 at 2.3 km/h.
ARTICLE IN PRESS
334
groups of 3 notches
2, 10, 4mm
880
860
840
2mm notch
A/D Conversion
820
4mm notch
6mm notch
10mm notch
800
780
760
740
15mm
10mm spacing
5mm
spacing spacing
720
700
680
660
640
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
Time / s
Fig. 14. Raw and normalised ltered data plots for test rails #5 and #3 at 32 km/h.
A/D Conversion
820
800
6mm notch
780
10mm notch
15mm notch
760
740
720
700
680
660
0.00
0.05
0.10
Time / s
0.15
0.20
Fig. 15. Raw and normalised ltered data plots for test rails #8 at 32 km/h.
7. Conclusions
High-speed ACFM inspection tests were carried out using a
micro-pencil ACFM probe that was driven by a TSC Amigo crack
microgauge system. Tests were carried out initially using a test
piece containing spark-eroded notches which were 2 and 4 mm
deep. A turning lathe was used to rotate the rotary test piece at
speeds up to 121.5 km/h (at 3000 rpm). This experimental
conguration provided controlled inspection conditions as well
as easy repeatability of the tests. It was found that the probe could
successfully detect the induced notches at 121.5 km/h and that the
ACFM signal remained largely unaffected by the increases in
ARTICLE IN PRESS
M.Ph. Papaelias et al. / NDT&E International 42 (2009) 328335
Acknowledgements
The provision of facilities and funding by the Centre for Rail
Research and Education and the School of Engineering at the
University of Birmingham is gratefully acknowledged. The authors
also wish to express their gratitude to TSC Inspections Systems
Ltd. for the provision of the ACFM equipment. Finally, the authors
would like to thank Dr Y. Fan at the Physics Department of
Warwick University for designing the trolley for the rig experiments and Mr Warren Hay and Mr Andy Dunn for their kind
assistance during testing.
References
[1] Ireland RC, Torres CR. Limitations of the circumferential MFL technique in the
NDE of pipelines. In: Proceedings of magnetics on non-destructive testing
Seminar. London, UK, April, 2005.
335