The University of Manchester Research
Surface-Breaking Flaw Detection in Mild Steel Welds usingQuantum Well Hall Effect Sensor Devices
Document VersionFinal published version
Link to publication record in Manchester Research Explorer
Citation for published version (APA):Watson, J. M., Liang, C-W., Sexton, J., Biruu, F., & Missous, M. (2018). Surface-Breaking Flaw Detection in MildSteel Welds using Quantum Well Hall Effect Sensor Devices. Paper presented at The 45th Annual Review ofProgress in QNDE Conference, Burlington, United States.
Citing this paperPlease note that where the full-text provided on Manchester Research Explorer is the Author Accepted Manuscriptor Proof version this may differ from the final Published version. If citing, it is advised that you check and use thepublisher's definitive version.
General rightsCopyright and moral rights for the publications made accessible in the Research Explorer are retained by theauthors and/or other copyright owners and it is a condition of accessing publications that users recognise andabide by the legal requirements associated with these rights.
Takedown policyIf you believe that this document breaches copyright please refer to the University of Manchester’s TakedownProcedures [http://man.ac.uk/04Y6Bo] or contact [email protected] providingrelevant details, so we can investigate your claim.
Download date:14. Apr. 2021
Surface-Breaking Flaw Detection in Mild Steel Welds using
Quantum Well Hall Effect Sensor Devices
James M. Watson 1 a), Chen-Wei Liang 1, James Sexton 1, Firew A. Biruu 1 and
Mohamed Missous 1 b)
1 The University of Manchester, School of Electrical and Electronic Engineering, Manchester, Greater Manchester,
M13 9PL, United Kingdom
Abstract. With growing industrial interest in Non-Destructive Evaluation (NDE) applications of Quantum Well Hall
Effect (QWHE) sensors, a study was conducted to quantitatively determine the detection capabilities and general
performances of low frequency QWHE imaging for surface-breaking flaws and their comparison with Magnetic Particle
Inspection (MPI), Eddy Current Testing (ECT) and Alternating Current Field Measurement (ACFM). In this study, a
probe consisting of a QWHE sensor, illuminating electromagnet and sensor circuitry was controlled using an automated
XY scanner with a measurement step of 250 µm to simulate an integrated array of QWHE sensors of 250 µm pitch. This
probe applied a 3 mT, 100 Hz frequency magnetic field to map the surface magnetic field and Magnetic Flux Leakage
(MFL) response of five bespoke dressed mild steel weld samples. These samples provided 15 surface-breaking flaws of
varying length from 3 to 11 mm; mainly longitudinal toe and center-line cracks which are representative of typical
requirements of our industrial partners. The same samples were also subjected to MPI, ECT and ACFM provided by
leading industrial companies using their own qualified personnel, equipment and procedures. The outcomes and
performance of each NDE technique including QWHE imaging were then compared and evaluated.
INTRODUCTION
An initial study was conducted to quantitatively determine the general performance and detection capabilities of
QWHE sensor-based magnetic imaging to compare them to existing, well established electromagnetic NDE
techniques. Samples of mild steel were synthesized by Sonaspection Ltd. with dressed welds and artificial flaws
before being subjected to MPI, ECT, ACFM and QWHE imaging inspections. These tests were performed on a
double-blind trial basis with each technique not being fully optimized to achieve “best” results. MPI was performed
by BAE Systems using a standard 50 Hz ~1.2 T yoke, contrast paint and kerosene-based media spray according to
their standard NDE procedures. ECT was performed by Rolls-Royce Nuclear with probes and inspection frequencies
available that were not optimized for mild steel welds. These ECT probes were mounted on an XY measurement
system to obtain limited imaging capabilities. ACFM was performed manually by TSC Inspection Systems using
their standard AMIGO probe, more suited to monitoring gross-sized flaws in service structures over small flaws of
this study. Their twin-field probes optimized for weld inspections were not available at the time of this study.
QWHE imaging was achieved using a probe consisting of a single QWHE sensor, sensor circuitry and an
illuminating electromagnet which applied a 3 mT, 100 Hz magnetic field. The XY movement of the probe was
controlled using a scanner of step size 250 µm to simulate an integrated QWHE sensor array of 250 µm pitch. A
significantly smaller step size of < 100 µm could have been used to obtain ever higher resolution images but would
not have been appropriate for these studies and would not represent a meaningful sensor array for millimeter size
NDE crack detection purposes. Throughout the QWHE imaging tests, an applied magnetic field frequency of 100
Hz was used. To date, there has been no research into an optimal inspection frequency or measurement steps for any
applications of QWHE imaging, including NDE crack detection.
The general performances, including main advantages and limitations of each technique are discussed in their
retrospective sections. The outcomes of each NDE inspection trial were also compared, evaluating:
The amount of flaws detected (not probability of detection).
Any additional indications.
The approximate indication signal amplitude to background floor level.
The accuracy of flaw sizing.
The approximate inspection times.
Samples
Fifteen flaws representative of those encountered in carbon-steel butt-welds were manufactured by Sonaspection
Ltd and distributed over 5 mild steel weld samples were used in this study. The flaws were predominantly center-
line and toe cracks of various lengths ranging from 3 mm to 11 mm, longitudinal to the direction of the welds, as
shown in Figures 1(a) to 1(e) below. Each sample measured 150 mm x 150 mm with a thickness of 20 mm and each
containing a full penetration weld with caps and roots ground flush. The parent metal was a generic mild steel with
electrical and magnetic properties comparable to those o AISI 1010.
(a) (b)
(c) (d)
(e)
FIGURE 1. Official validated sketches provided by Sonaspection Ltd. of each sample and used in this study. Flaws present on
cap side unless otherwise stated. (a) Sample 6441-01 with 11 mm toe crack, 13 mm linear porosity (root side) and 10 mm toe
crack. (b) Sample 6441-02 with 8 mm toe crack, 8 mm center-line crack (root side) and 8 mm toe crack. (c) Sample 6441-03 with
5 mm toe crack, 5 mm linear porosity (root side) and 5 mm toe crack. (d) Sample 6441-04 with 5 mm toe crack, 4 mm center-line
crack (root side) and 5 mm toe crack. (e) Sample 6441-05 with 3 mm toe crack, 3 mm center-line crack (root side) and 3 mm toe
crack.
Magnetic Particle Inspection
Arguably the easiest and most common NDE technique for ferromagnetic materials, MPI is a fast and efficient
method particularly for inspecting large surface areas whilst providing instantaneous results as an image [1]. It has
low start-up and running costs as well as being comparatively easy to use and qualify in industry. In practice, it
suffers from a low detection limit of ~3 mm and a poor probability of detection, depending on the specific industrial
application [2]. Unlike other electromagnetic NDE methods, it cannot provide any flaw depth measurements or
estimates, its flaw length measurements are inaccurate with false indications being common, particularly when
sample geometry acts as a natural MFL site (e.g. filet welds). It requires surface preparation, including pre-cleaning
and is invasive as coatings must be removed [3]. It is also a messy technique which typically requires post-test
cleaning and can contaminate / disrupt some production line stages.
MPI was performed on the samples (Figures 2(a) and 2(b)) by BAE Systems using their industry standard
approved 50 Hz ~1.2 T yoke, standard white contrast paint and kerosene-based black magnetic ink spray media. A
3-lines “Burmah-Castrol strip” magnetic flux indicator was used to validate the strength and performance of the
equipment, as standard to their procedures.
(a) (b)
FIGURE 2. Photographs of MPI indications of two flaws from this study. (a) MPI indication of 4 mm center-line crack on
sample 6441-04 root side. (b) MPI indication of 5 mm linear porosity cluster on sample 6441-03 root side.
These MPI tests were able to detect all of the flaws with no false indications, as well as being able to size and
position them within 1 mm error. It must be stated that these were manufactured as MPI samples, meaning that the
flaws were designed to provide the most realistic MPI indications and the “correct” positions and lengths were
validated by Sonaspection Ltd.’s own MPI tests. Consequently, the indications obtained by other methods may not
be as clear or as realistic and are subject to some reasonable tolerance.
Eddy Current Testing
Although ECT is a group of different electromagnetic NDE techniques, in this study only conventional eddy
current testing was used (i.e. single frequency, absolute, phase and amplitude measurements). As such, it is this
method which ECT will be referred as. ECT is known to be a very sensitive electromagnetic NDE technique where
modern equipment and software allow inspections to be set-up and performed easily. It also provides instantaneous
results usually through the real-time plotting of a complex impedance plane [4]. Compared to other techniques, it
has a good detection limit and probability of detection in practice [2], with a sensitivity enabling reliable flaw sizing
estimates based on pre-test calibrations [5]. In most applications, ECT is non-invasive meaning non-conducting
coatings do not have to be removed, depending on the coating thickness, sample material properties and test
parameters. However, flaw characterization is difficult with ECT, particularly when using complex impedance plane
analysis. As with other sensor-based electromagnetic NDE methods, it is sensitive to lift-off and sample geometry,
the main drawback being a comparatively slow inspection time based on the small footprint of ECT probes [6]. This
makes it ineffective at inspecting large surface areas, with false indications common in practice due to its sensitivity.
Rolls-Royce conducted the ECT trials using Zetec absolute probes of frequency 2.5 MHz and 6 MHz, both
probes and frequencies were not necessarily optimized for mild steel weld inspection. A standard XY measurement
stage was used to provide limited imaging capabilities (Figures 3(a) and 3(b)) using Magnifi software. A standard
ECT calibration block was used to provide flaw depth estimates.
(a) (b)
FIGURE 3. ECT phase shift images of complete weld scans of two samples from this study. (a) ECT indications of two 8 mm toe
cracks on sample 6441-02. (b) ECT indications of 5 mm linear porosity cluster on sample 6441-03.
Unfortunately, only 4 of the samples were tested due to time constraints. However, from the samples tested all of
the flaws were detected, including the smallest 3 mm cracks. This method did produce 4 additional indications
relating to microstructure, 2 with comparable magnitude to flaw signals, the other 2 with lower magnitudes. The
noise from one such microstructure, suspected to be lack of fusion, is shown in Figure 3(a). ECT was able to size
63% of flaws within 1 mm error, the remainder within 2 mm error. Likewise, 82% were positioned within 1 mm
error, the remainder within 2 mm error.
Alternating Current Field Measurement
ACFM is a strongly growing and emerging electromagnetic NDE technique that is mainly used for the detection
and accurate sizing of gross flaws of in-service structures, typically in inaccessible places such as offshore rigging
[7]. Most notably, compared to other techniques, ACFM equipment has been fundamentally designed with the end
user in mind. Consequently, ACFM inspections are very easy to set-up and perform, with instantaneous results in
real time through the ASSIST software. In practice, it is very accurate at sizing flaw lengths and depths using the
ASSIST algorithms and indication database. There is also a large coverage from a single probe (compared to
conventional ECT) with inspection frequencies used that require none/minimal surface preparation. This technique
is non-invasive as non-conductive coatings do not have to be removed [8], dependent on the sample material
properties and coating thickness. However, as with other sensor-based electromagnetic NDE methods, ACFM is
sensitive to lift-off and sample geometry; and can this can add to difficulties in characterizing flaws. The AMIGO
probes have a poor detection limit as their main applications are for gross flaw detection and accurate sizing. As
such, it must be noted that the ACFM technique could be optimized to have a much better detection limit, with
limited industrial/commercial demand preventing this.
For this study, ACFM was performed by TSC Inspection Systems, the leading manufacturer of ACFM
equipment, using their standard AMIGO probes which were not optimized for mild steel welds. Manual
measurements were taken using a standard 50 kHz AMIGO probe (Figure 4(a)) with magnetic field component
measurement plots and flaw sizing estimates (Figure 4(b)) autonomously compiled into a report after the testing was
complete.
(a) (b)
FIGURE 4. (a) ACFM manual measurement lines drawn on sample 6441-02 with flaw measurements. (b) ACFM indications of a
11 mm toe crack on sample 6441-01.
As expected, 3 of the smallest flaws (one 4 mm and two 3 mm cracks) were undetected by the ACFM AMIGO
probe, as the typical applications are much different. However, 90% of the flaws detected were sized correctly or
within 1 mm of error, the remainder within 2 mm error. None of the flaws were positioned correctly due to the
measurements being entirely manual. However, they were all within 2 mm of error.
Quantum Well Hall Effect Sensor Imaging
Developed and tested by The University of Manchester, an XY scanner and QWHE sensor probe (Figure 5(a))
were used to apply a 3 mT, 100 Hz magnetic field to each sample under test, using a QWHE sensor to measure the
MFL [9] caused by each flaw (Figure 5(b)). The effective active sensing area of the QWHE sensor was 160 x 160
µm2.
(a) (b)
FIGURE 5. QWHE sensor probe used in this study. (a) Photograph of probe taken whilst scanning sample 6441-03. (b) Diagram
of probe showing general layout of components.
A standard depth of penetration of 0.6 mm was approximated based on the sample material properties and
inspection frequency of 100 Hz. A measurement step of 250 µm was used which simulates a representative
integrated QWHE sensor array of 250 µm pitch. The differential of the magnetic field was then calculated and
normalized against the measurement step by calculating the magnetic field gradient. These steps are mathematically
described in Equations (1) and (2), where 𝑩𝑥𝑦 represents the magnetic field measurement taken at coordinate (𝑥, 𝑦),
with a measurement step of 𝑑.
𝑩′𝑥𝑦 = 𝑩𝑥𝑦 − 𝑩(𝑥−𝑑)𝑦 (1)
𝑨′𝑥𝑦 = 𝑩′𝑥𝑦
𝑑 (2)
This value 𝑨′𝑥𝑦 was mapped in real time whilst the scan was ongoing and was plotted to produce QWHE MFL
images of each sample to be used in this study. An example of these images is given as Figure 6(a), along with
annotations explaining the various features and artefacts in Figure 6(b).
(a)
(b)
FIGURE 6. QWHE images of sample 6441-01. (a) Unprocessed QWHE image of sample 6441-01 in greyscale. The two toe
cracks appear as bright white indications within the weld boundary line. (b) Annotated version of Figure 6(a) highlighting the
indications of the toe cracks and other weld features. Blue rectangle is used in Figure 8.
A basic hill shading image processing technique was used to identify further areas of research (i.e.
microstructure). This technique visually identified the contours produced by all the MFL responses of each sample
by creating shadow-like artefacts on each image. An example of this is shown in Figure 7.
FIGURE 7. Processed QWHE image of sample 6441-01 in greyscale using basic 45° hill shading to highlight microstructural
information contained within QWHE images.
Figures 6 and 7 show that low frequency QWHE MFL imaging can not only detect the presence of representative
NDE flaws within mild steel welds but is also sensitive to microstructural changes using a 100 Hz inspection
frequency. As such, an intensity plot (Figure 8) is given showing the relative intensity differences between the MFL
responses from a crack and those from microstructures.
FIGURE 8. An intensity plot showing the relative magnitudes of MFL response from a crack and microstructures from sample
6441-01.
Figure 8 was created by compiling the data within the blue rectangle slice on Figure 6(b) onto a single intensity
vs. position graph. It shows that the intensity from MFL responses of cracks are far larger than those of
microstructures, emphasizing that QWHE sensors are not “over-sensitive” for NDE crack detection applications.
This promising initial work suggests that gross microstructural changes (such as lack of weld fusion) give maximum
MFL responses that are 1
3 the magnitude of cracks, giving this specific technique and test parameters an approximate
Signal to Noise Ratio (SNR) of 3 (~4.8 dB) at this stage.
As with the other NDE techniques employed in this study, QWHE imaging was not optimized for crack
detection in mild steel welds as the effects of numerous inspection parameters have yet to be explored, determined
and hence optimized. These include but are not limited to: the frequency of the applied magnetic field (i.e.
inspection frequency) and illumination set-up (electromagnet and/or coils, their design, etc.). As previously
mentioned, the measurement step of 250 µm used simulated an integrated QWHE sensor array of 250 µm pitch,
representative of realistic future applications of QWHE NDE sensor technology [10][11][12]. A measurement step
of < 100 µm could have been used to obtain “better” higher resolution results but would not have best represented
the future aims of technology (larger pitch arrays needed to maximize inspection area / footprint).
QWHE MFL images were obtained for each sample’s cap and root side. Using these, all of the flaws were
detected using the unprocessed images, similar of those in Figure 6. Several additional indications were found and
are believed to be MFL responses from microstructural variations within the samples, including lack of weld fusion.
As shown by Figure 8, these additional indications had magnitudes far below the magnitudes of MFL from a crack
and created a noise floor from which the MFL from cracks (and the MFL from the flaw coupon boundaries) were
clearly visible. In this trial, it was found that using the unprocessed images grossly undersized the flaw lengths.
However, the basic 45° hill shading image processing technique was able to size all the flaw lengths within 2 mm of
error which is a promising start to future research in flaw profiling using QWHE images. The basic XY scanner used
in this study was unable to provide reliable positioning.
Overview
As expected, MPI appeared to perform more favorably based on the nature of the samples, how the sample
sketches and flaw sizes/positions were validated, as well as the sheer versatility and performance of the technique in
controlled settings. Also, since the ACFM was performed manually using an AMIGO probe which was designed for
completely different industrial applications, these results are not representative of the optimized performance of
ACFM for small crack detection in mild steel welds.
However, Table 1 shows that in this initial study QWHE imaging performed comparably well to conventional
ECT in this controlled setting. The additional indications from microstructure can be ignored using future post-
processing techniques and a more optimized applied magnetic field frequency (i.e. inspection frequency) at least for
these types of inspections; in others, microstructures would be prime targets as in large grain oriented electrical steel
materials.
TABLE 1. An overview of the performance of each NDE technique used in this study.
MPI ECT ACFM QWHE
Flaws found 100% 100% 100% 100%
False / additional
indications 0 4 1 9
Indication signal
amplitude to
background ratio
n/a ~4.8 dB
(SNR of ~3)
~4.8 dB
(SNR of ~3)
~4.8 dB
(SNR of ~3)
Flaw length sizing
accuracy ± 1 mm ± 2 mm ± 2 mm ± 2 mm
Total inspection
time
~ 1 hour
includes pre and
post cleaning
~ 6 hours ~ 4 hours
~ 6 hours
based on an
integrated 8 x 1
250 µm sensor
array – could be
further optimized
Notes
Official sample
drawings based
on manufacturer’s
MPI validation
tests
2 microstructural
additional
indications of
similar intensity
to flaw signal
AMIGO probe
not designed for
flaw detection of
this scale
All microstructural
additional
indications have
much smaller
intensity than flaws
It is believed that there are several near-surface flaws (within 1 mm of surface) which have been detected using
QWHE imaging at 100 Hz but not detected by ECT or ACFM due to their much smaller depths of penetration based
on their much higher inspection frequencies. The limited sensitivity of MPI would also explain why these were not
detected in the MPI tests performed by the sample manufacturers or as part of this study. In addition, since these
samples were manufactured as MPI samples, any validating ultrasonic NDE performed would have deemed these
flaws negligible for MPI use. As such, in the near future Ultrasonic Phased Array (Total Focusing Method), other
available NDE methods and finally destructive techniques will be performed to further investigate these claims of
microstructural anomalies and sub-surface flaws.
Conclusions
From this initial study, the detection capabilities and general performance of MPI, ECT and ACFM were
compared to low frequency QWHE MFL imaging, for the specific application of surface-breaking flaws in dressed
mild steel welds. This study showed that QWHE imaging can be used to detect surface-breaking flaws and also
evaluate microstructures. It was found that in this controlled setting, QWHE imaging performed comparably well to
ECT in terms of detection capability and SNR, with QWHE sensors providing full imaging capabilities technology
[10].
Future work as part of this research will be focused on developing the maturity of the QWHE imaging technique:
Frequency optimization for crack detection in mild steel welds.
Measurement step (sensor array pitch) optimization for crack detection with largest inspection area possible.
Development of image processing and enhancement techniques.
Development of automated flaw detection algorithms.
Development of database of MFL responses from cracks of known dimensions for flaw reconstruction (flaw
length and depth measurements).
Development of multi-frequency / pulsed applied magnetic field technologies for QWHE imaging for
different materials and NDE applications (such as corrosion mapping).
ACKNOWLEDGEMENTS
Industrial involvement of BAE Systems Submarines, Rolls-Royce Nuclear and TSC Inspection Systems for
conducting their parts of this trial.
EPSRC-EP/P006973/1 “FUTURE COMPOUND SEMI-CONDUCTOR MANUFACTURING HUB”)
EPSRC-EP/LO22125/1 “UK RESEARCH CENTRE IN NON-DESTRUCTIVE EVALUATION(RCNDE3)”
iCase PhD Studentship through RCNDE (BAE Systems Submarines)
REFERENCES
1. British Standard BS EN ISO 9934-1:2016. Non-destructive testing – Magnetic particle testing. Part 1: General
principles (ISO 9934-1:2016).
2. British Standard BS 7910:2013+A1:2015. Guide to methods for assessing the acceptability of flaws in metallic
structures.
3. Burke, S. K., 2013. Review of Literature on Probability of Detection for Magnetic Particle Nondestructive
Testing. Department of Defence, Australia.
4. British Standard BS EN ISO 17643:2015. Non-destructive examination of welds – Eddy current examination of
welds by complex plane analysis.
5. British Standard BS EN ISO 15549:2010. Non-destructive testing – Eddy current testing – General principles.
6. Udpa, S. S., 2004. NDT Handbook – Electromagnetic Testing (3rd ed.). Columbus: American Society for
Nondestructive Testing.
7. Lugg, M. C., 2011. Applications of ACFM for Weld Inspection by ROV. Singapore International NDT
Conference & Exhibition , 3-4 November 2011.
8. TSC Inspection Systems, 2016. AMIGO Specifications Sheet. http://www.tscndt.com/wp-
content/uploads/2017/12/specifications-sheet-amigo.pdf.
9. Rao, B. P. C., 2012. Magnetic Flux Leakage Technique: Basics. Journal of Nondestructive Testing &
Evaluation, 11(3), 7–17.
10. Liang, C. W., Ahmad, E., Balaban, E., Biruu, F. A., Sexton, J., & Missous, M., 2017. A Real Time Quantum
Well Hall Effect 2D Handheld Magnetovision System for Ferromagnetic and Non-Ferromagnetic Materials
Nondestructive Testing. 56th Annual Conference of the British Institute of Non-Destructive Testing, BINDT
2017.
11. Biruu, F. A., Balaban, E., Ahmad, E. and Missous, M. Quantum Well Hall-Effect Sensor Based Handheld
Magnetic Scanner with Programmable Electromagnetic Coil for Non-destructive Testing of Ferromagnetic and
Non-ferromagnetic Materials. 55th Annual Conference of the British Institute of Non-Destructive Testing,
BINDT 2016.
12. J. M. Watson; C.W. Liang, E. Ahmad, J. Sexton; M. Missous (2017) Surface Crack Detection in Dressed Steel
Welds Using Advanced Quantum Well Hall Effect Sensors: 56th Annual Conference of the British Institute of
Non-Destructive Testing, BINDT 2017.