Channa Nageswaran and Anaϊs Bourgeon
TWI Ltd, Granta Park, Great Abington, Cambridge, CB21 6AL, UK
OMS, Twyford Business Centre, London Road, Bishops Stortford, CM23 3YT, UK
Paper presented at INSIGHT 2012
This paper presents the work done towards developing inspection methods for application on partially completed girth welds in metallurgically clad pipes. Being able to detect flaws prior to the deposition of the fill runs provide key technical and commercial advantages. Phased array techniques and a proprietary laser optical system implemented from inside the pipes were evaluated for inspection of critical weld flaws. This paper shows that inspection solutions readily available to industry at present can aid in improving the welding of clad pipeline girth welds.
The inspection solution is a key aspect of laying good quality pipelines in the oil and gas industry. The use of corrosion resistant alloy clad pipelines is a necessity to transport the increasingly corrosive fluids being recovered around the world. Any flaws found at the intermediate weld deposition stage can be repaired with much less risk to the integrity of the joint when compared to having to do the repair once the weld is complete.
The aim of the work was to develop and demonstrate reliable weld inspection techniques of partially completed girth welds in metallurgically clad pipes. The techniques developed were envisaged to be deployed at station 2 on a lay barge, replacing the radiography techniques currently being applied at this welding stage. However, radiography is less sensitive to some flaw types than ultrasonics 1, whereas ultrasonic techniques can suffer from the interfaces between the parent, cladding and weld, in addition to the geometry at the root to hot pass region of the partially deposited girth weld 2.
In addition to the use of ultrasonic techniques from the outside surface of the pipe to interrogate the volume, a laser optical system was investigated for implementation from the inside surface of the pipe to inspect the surface condition of the root.
The work presented in this paper was undertaken as part of a project funded by several Sponsors in the oil and gas industry.
2.1 Specimen materials and manufacture
Four development specimens were manufactured. The specimens were 14" outside diameter (OD), 13mm thick X65 steel pipe metallurgically clad using 3mm of 316L stainless steel and the welds were made using 2209 grade Duplex steel filler. The following types of simulated flaws were artificially introduced into the specimens which were welded at the root and hot pass only:
- Porosity (root and hot pass)
- Lack of root penetration (without misalignment)
- Lack of root penetration (with misalignment, 0.5mm)
- Lack of side wall fusion on the hot pass (>2mm)
The specimens were examined by radiography to confirm the presence of the implanted flaws and three of the specimens were sectioned subsequent to the development work to confirm the size of the implanted flaws and compare with the results of the inspection techniques.
Figure 1 shows the remaining groove from the partially completed weld, ie deposition was halted after the root and hot pass. The artificial flaws implanted in one half of each specimen pipe, represented the type of flaws expected at this intermediate welding stage and hence the ultrasonic techniques were required to detect them with confidence, providing both flaw type identification and accurate sizing information.
Figure 1 Image of specimens used in the project, highlighting a typical weld groove where the fill passes were not introduced to complete the weld
Figure 2 Average measured profile of the partially completed specimens.
The incomplete fusion in the hot pass (specimen 08) was simulated by introducing a machined slot along the fusion boundary and the flawed area could be visually detected from the outside surface. The presence of porosity (specimen 01) was confirmed through radiography and the presence of the lack of penetration type flaws (specimens 04 and 06) could, to a limited extent, be confirmed through tactile investigation of the weld root. Note that specimen 06 was welded with deliberate misalignment, whereas there was (nominally) no misalignment in the other pipes. In each pipe specimen four flaws were introduced, equally spaced in the region from 0 to 180° around the circumference of the weld. This allowed for half of the weld circumference available for use as an ultrasonic reference and to remove samples for metallurgical analysis (see Section 4).
2.2.1 Phased array system
Phased array technology is characterised by the use of array transducers and instrumentation (termed array controllers) that are able to address each element of the array independently of other elements. The array controller is then able to fire or generate a sound wave from each element at different times with respect to other elements on the array. The relative times at which the elements are fired and the subsequent wave fronts emanating from them propagate into the medium ahead where they interfere with each other, in effect allowing a measure of control over the resultant sound field. This control allows the array probe to be more versatile compared with conventional single element ultrasonic probes, where the sound field is fixed, so it is possible to better tailor the ultrasonic techniques for the inspection application 3.
Following basic ultrasonic tests using conventional systems (including single element contact probes and immersion scanning) to study attenuation and geometric limitations, a high frequency system was specified for inspection of the 14" OD components. Phased array probes and instrumentation were elected for implementing a pulse-echo configuration for the inspection. The techniques were designed and the parameters were optimised through simulation, in particular the specification of the array probes.
The phased array techniques were implemented using two array controllers: Zetec 4 DYNARAY and Olympus NDT 5 OMNISCAN. The DYNARAY was used for the development as it provided the better capability in terms of signal-to-noise (S/N) performance, pulser energy and electronic filters, which allowed for investigating and developing the most suitable techniques. However, the techniques were designed for field application at station 2 on a lay barge, for which the smaller, lighter, more ruggedized OMNISCAN was better suited.
Optimisation studies at frequencies from 2 to 15MHz were undertaken in pulse-echo mode using both longitudinal and shear waves, with the criteria being attenuation and back scattered S/N performance. Given the restricted geometric condition of the incomplete weld, a balance between the optimisation criteria to resolve a small flaw within the geometry was found at a frequency of 10MHz. A standard Olympus NDT 10L32-A1 probe was chosen to implement pulse-echo shear and longitudinal beams into the specimen directed towards the root and hot pass. The probe was a 32 element linear array with a pitch of 0.31mm and an element width of 7mm. A standard sector scan range of beams from 45 to 70° at an increment of 0.5° were focused to the depth of the root, i.e. 16.5mm, through a specifically designed Rexolite wedge. Scanning sensitivity was chosen with reference to the sensitivity to set the corner echo from the cladding at the end of the pipe for each beam, which captured the specific attenuative quality of the cladding.
To automate the inspection, the MAGSCAN developed by Phoenix ISL 6 was selected. The scanner is placed on the pipe and remains attached through its magnetic wheels with the probe attached to a scanning arm; Figure 3 shows the scanner mounted on specimen 01 (containing porosity).
Figure 3 The 10MHz array probe connected to the scanning arm of the MAGSCAN and positioned to implement a phased array technique.
2.2.2 Laser optical system
Optical Metrology Services (OMS) 7 manufacture and operate a number of proprietary laser and optical weld scanning systems for out-of-roundness of pipes and internal weld scanning of girth welds in pipes. The OMS Weld Scanner system was used to sentence the welds to the criteria outlined in DNV-OS-F101 8 for visual examination. In addition to optical data, the OMS system has a laser measurement dimensional accuracy of 0.05mm in a radial direction from the pipe centre and the system takes linear laser stripe scans at 2.85° intervals. Figure 4 shows the Weld Scanner being used to inspect the internal condition of the root in a partially girth welded specimen.
Figure 4 The OMS Weld Scanner inspecting the inside root condition.
3.1 Phased array system
The phased array scan data is presented using methods and conventions established by Olympus NDT. Figure 5 shows a typical data display for analysis of the weld condition with the interlinked cursers:
- The sector scan (termed the S-scan) in the right hand data window is composed of all the beam angles generated by the array; there is a sector scan for each 1mm point around the circumference. On the S-scan, a pink outline of the weld has been drawn (the overlay) to aid interpretation of the signals; the incomplete nature of the weld is represented by the horizontal line on the overlay 8mm from the root above which no weld material was deposited. An echo due to a flaw is displayed on the cross-section of the weld overlay and it’s through-wall extent can be evaluated. The black cursor line running through the sector selects the beam angle of interest which is used to display the B- and A-scan views.
- The B-scan view (the bottom left hand data window) displays the echoes generated along the black cursor line in the S-scan view; thus the echoes along a certain beam angle which could, for example, be incident in the root region can be viewed in the B-scan view. If there was a flaw in the root this would be displayed in the B-scan view showing its circumferential extent; gates remove irrelevant data such that the B-scan view mimics a strip-chart view around the circumference. The S-scan shown is selected by the vertical green cursor shown in the B-scan view.
- The A-scan view (the top left hand data window) displays the fundamental ultrasonic time-amplitude data from which both the S- and the B-scan data presentations are generated. The A-scan shown is from the beam indicated by the black cursor in the S-scan view. The signal amplitude due to a flaw and the S/N condition of the inspection can be evaluated using the A-scan view.
Figure 5 Data display showing the three views for the A-, B- and S-scans along with the two cursors (black and green) which interlink all the data between these three views.
The quality of the data is based on good coupling, lack of missing circumferential data, good positioning of the wedge with respect to the weld centreline so that the weld overlay position is accurate with respect to the data and good S/N performance in the A-scan view. It would be the responsibility of the site team to ensure that the data quality is sufficient (as specified in field implementation procedures) before the weld is analysed and sentenced.
In Figure 5, in the S-scan view, the signal marked A is a geometric signal due to the back of the root; often termed the root signal in a conventional ultrasonic technique and runs the full length of the weld. The signal marked B is also a geometric signal due to the root profile, as is the signal marked C which is due to the lower beam angle reflecting on the back surface (marked as the horizontal blue line at a depth of 16.5mm in the S-scan) and being incident on the edge of the groove at the root. Both A and B are signals which will be present in fully completed welds but C, plotting at a depth of 25mm, is unique to partially completed welds.
In Figure 5 there are two features marked D and E in the B-scan view due to two incomplete fusion type flaws in specimen 08. Note that the vertical green cursor is in a region around the circumference where neither flaw D or E is present; hence the cross-section shown by the S-scan view in Figure 5 does not show either of the flaws, but only the nominal geometry signals. In Figure 6, the green cursor cuts through the indication marked D and now the S-scan in Figure 6 shows a cross-section through the weld at that circumferential point. It is then straight forward to identify the flaw (ie detect it), class it as incomplete fusion due to its position on the weld bevel and further evaluate its size based on its position on the weld bevel. The image of the flaw cross-section given by the signal marked G (below the back wall marked by the horizontal blue line) allows the operator to confirm and check the size measured directly using signal marked F above the back wall (horizontal blue line). The circumferential extent of the flaw can be measured in the B-scan view.
Figure 6 Data through the incomplete fusion flaw marked D; the S-scan shows a cross-section through the weld at the circumferential point (green cursor) and the A-scan view shows the time-amplitude signal (black cursor).
Figure 7 (a) shows the presence of implanted porosity in specimen 01 using radiography and Figure 7 (b) shows the ultrasonic data in the B-scan view.
Figure 7 Radiographic and ultrasonic data showing the presence of porosity in specimen 01.
Similarly, lack of root penetration flaws in specimen 04 could be detected, positioned and sized as shown in Figure 8. Note that the data in Figure 8 was collected using the OMNISCAN by an independent operator as part of blind trials using the procedure developed using the DYNARAY (Figures 5, 6 and 7). The operator was able to position, size and classify the flaw.
Figure 8 Data showing a lack of root penetration in specimen 04, which was reported to start at 993mm from datum, with a length of 26.5mm, with ligament linked to root and a maximum through-wall height of 3.6mm.
3.2 Laser optical system
Figure 9 shows the data output from the laser optical system which shows the measured root profile and an optical image of the region. The recorded misalignment in specimen 06 taken directly from the sample using a weld gauge was 2.5mm; the OMS system measured the misalignment every 1mm around the pipe with five hundred measurements across the weld, measuring a maximum misalignment in this sample of 2.61mm.
Figure 9 Laser optical measurement of a lack of root penetration in specimen 04 with a measured length of 10mm, which nominally had no misalignment. The left image shows a typical laser measured root profile to evaluate misalignment and the right image shows the presence of a lack of root penetration type flaw in the weld.
The data from the OMS Weld Scanner and the phased array technique for length measurement of lack of root penetration in specimen 06 is given in Table 1, which shows reasonable correlation between the two data sets in flaw length sizing, with a maximum discrepancy of 8mm in the first flaw.
Table 1 OMS data showing length and relative position of flaws in specimen 06 compared with the phased array data.
||Phased array data
|Flaw position (mm)
||Flaw length (mm)
||Distance to start of next flaw (mm)
|Flaw position (mm)
||Flaw length (mm)
||590 - 610
||658 - 703
||750 - 810
||870 - 910
4.1 Sizing capabilities
Selected areas of specimens 01, 04 and 08 were sectioned to investigate the through-wall sizing capability of the phased array techniques. The locations chosen were based on the scan data, ie locations where flaws were identified. Figure 10 shows the general weld cross-section through specimen 01 through a region of porosity marked C as shown in Figure 11. Two pores are visible in the micrograph of Figure 10 in the hot pass. However, note also the lack of side wall fusion in Figure 10 which is also evident in the data of Figure 11 at higher beam angles in the sector scan.
Figure 10 Cross-section through a point in the specimen corresponding to the section in Figure 11 marked C, showing pores and lack of fusion on the hot pass side wall.
Figure 11 Data showing flawed regions containing porosity in specimen 01, along with a lack of side wall fusion in the hot pass.
It is only possible to size regions containing porosity along their circumferential extent which was confirmed through the radiography; as the volume of the weld is small, through-wall sizing is not required and the weld containing porosity will be sentenced based on the circumferential extent of the damage.
Three lack of root penetration flaws were detected in specimen 04 using the phased array techniques, shown in Figure 12, marked A - C. Sections were taken through all three marked regions at 663mm (A), 907mm (B) and 992mm (C) from the datum.
Figure 12 Data showing flawed regions containing lack of root penetration in specimen 04.
Figure 13(a) shows the micrograph at 663mm showing the lack of root penetration flaw corresponding to the data shown in Figure 12. Similarly, Figures 13(b) and 13(c) show the micrographs corresponding to positions marked B and C in Figure 12.
Figure 13 Micrographs of the section in specimen 04 at (a) 663mm, (b) 907mm and (c) 992mm from the datum, focused on the lack of root penetration flaw (highlighted).
The incomplete fusion flaws implanted in specimen 08 could be measured from the surface of the pipe, visible to the naked eye, but these specimens were also sectioned.
The data was analysed independently by an EN 473 Level II Phased Array UT (CSWIP) operator to evaluate through-wall sizing accuracy against the sectioned data on lack of root penetration. The operator was made aware of the incomplete nature of the weld inspected and details such as dimensions, velocities, probe positions etc to be able to relate the sector scan to the component.
The operator was able to detect all but one of the indications present in specimens 04 and 08. If the specimens were to be rejected if any flaw-like indications were identified in the root region (based on circumferential length only) then the techniques should perform well in qualification trials. If the acceptance criteria depend on through-wall sizing (for lack of root penetration and incomplete fusion) then further qualification evidence is required to assess the statistical sizing accuracy of the techniques. The operator deemed it possible to size the indications but the sizing accuracy was not quantified in this project through the use of salami-sectioning (as specified in DNV OS-F101 8) as the flaw set was insufficient.
4.2 Influence of metallurgy on inspection capability
In general, it has been noted in previous work done at TWI and elsewhere that welds in the austenitic state and those which contain coarse grains are difficult to inspect ultrasonically because of the increased sound scatter and distortion of the sound field that can be induced by the material structure on the propagating sound wave. However, the experience of inspecting the development specimens was that the Duplex 2209 weld deposit (ie hot pass and root) and the 316L cladding did not present a significant challenge for penetration of the sound. One reason was that the volume of the metal in the region of interest was small but further evidence for the favourable material condition was evaluated, through metallurgical analysis using the Electron Back Scatter Diffraction (EBSD) technique 9.
A transverse metallographic section was extracted from the weld area in specimen 01 and mounted in conductive resin for analysis. The EBSD scan was carried out at a magnification of 50 times and the step size selected was 4μm (the distance that separates each point from its neighbour). Figure 14(a) shows the inverse pole figure map and the texture shows that the microstructures of the three areas, the duplex stainless steel weld, pipe steel and 316L cladding material, are very dissimilar. The pipe steel microstructure consisted of very fine grains, as expected, and a magnification of 50 times is not appropriate to assess the microstructure in more detail. The cladding material consisted of equiaxed grains of approximately 70μm diameter on average, consistent with the microstructure expected in 316L sheets. The microstructure in the weld was much coarser than in the other two areas. The elongated shape of the grains and their orientations are directly related to the direction of solidification of the weld metal (from the fusion line to the centre of the weld pool). Figure 14(b) shows the phase map which clearly discriminates between the different materials, hence the steel is fully ferritic, the cladding fully austenitic and the weld has a duplex structure (a mixture both phases).
Figure 14 EBSD scans showing (a) the texture differences between the weld, parent and cladding, and (b) the phase map where green is ferrite and red is austenite.
The wavelength of the shear wave at a frequency of 10MHz is about 300μm, which is much larger than the grains of the steel and is also an order of magnitude larger than the grains of the cladding. This goes some way to explain why the shear wave is able to penetrate the cladding, as the grains of the cladding are on the boundary of becoming significant scatterers of the sound; grains become significant scatterers when their size approaches 1/10th of the wavelength 10. However, by the same argument, the shear wave of wavelength ~300μm will be significantly scattered by the elongated grains in the weld metal. However, note that the angle at which the beams are incident into the weld implies that the wave is being affected by the grain width (similar in size to the cladding grains) rather than the longer lengths of the dendrites; hence the scattering in the weld is also on the 1/10th of the wavelength boundary and, as supported by the experimental evidence, could explain why the propagation is not adversely influenced.
The structure of both the cladding and the weld has an important effect on the capabilities of the ultrasonic techniques. In the case of the development specimens, which used the Duplex weld and the 316L cladding, the metallurgical analysis indicates that the ultrasonic waves (longitudinal and shear) should be able to penetrate the metals, sufficiently, to interrogate the entire weld and the experimental evidence, presented in this paper, supports this hypothesis.
The structure of the cladding is influenced by the method used to generate it, whether it was introduced through metallurgical bonding (such as roll bonding) or through fusion (weld-deposited) methods, such as strip-cladding. Metallurgical bonding does not lead to fusion and solidification that are known to coarsen the grain size of austenitic materials. In this project, only cladding applied by metallurgical bonding was investigated ultrasonically; however, there are occasions when weld-deposited cladding may be used. Thus, work was done to provide some guidance on the feasibility of using ultrasonics, in such cases, to detect critical flaws.
A further specimen was prepared for EBSD scanning, which was a stainless steel cladding, deposited using strip-welding (as opposed to the previous metallurgically bonded clad specimens). The scan was performed at a magnification of 50 times and the step size was 4μm. Figure 15 is an inverse pole figure map of the specimen which shows that, contrary to the cladding material that was metallurgically bonded, the weld-deposited clad material was strongly textured. The grain sizes are substantial with respect to both the longitudinal and shear waves, along both directions of their aspect, so propagating a wave through the deposited cladding will be much more difficult than with cladding introduced through metallurgical bonding. Work done by TWI and elsewhere support this conclusion, as the distortive and attenuative effects of cladding have been shown to seriously degrade inspection capabilities.
Figure 15 Inverse pole figure map from the scan of the weld-deposited cladding showing large dendritic grain structures.
The recommendation from the metallurgical analysis is that, wherever possible, the cladding should be applied using metallurgical bonding and the use of longitudinal or shear wave techniques should be feasible. If the cladding has to be deposited, then the design of the ultrasonic techniques will have to deal with higher attenuation (through scattering) and increased distortion, leading to complication in the interpretation of the signals. Note also that a metallurgically bonded pipe will often have a region of weld-deposited cladding when it has been made by the UOE process and, therefore, has a longitudinal seam, made by the SAW process and having the cladding made good by weld-deposition process, such as strip cladding. The inspection techniques should take account of the presence of this differential region as part of the interpretation procedure and, if deemed to be critical, then customised inspection techniques (possibly non-ultrasonic) may need to be specified for this particular region.
4.3 Further considerations
Inspection of the weld at the intermediate stage implies high temperature implementation, above 100°C. At these temperatures the use of water will be difficult but additionally consideration must be given to the contamination of the weld due to the exposed groove. An alternative solution would be to use high temperature couplant which, unlike water, will not evaporate at temperatures greater than 100°C. Additionally, the viscosity of the couplant is such that it may be more effectively prevented from flowing into the weld groove if contamination (either by water or couplant) is considered an issue.
Experiments were conducted to simulate high temperature inspection through a Rexolite material wedge. The criterion is to ensure that the temperature at the surface in touch with the probe crystal does not reach 90°C, which would lead to de-poling (degradation) of the piezoelectric active material of the probe. The results show that the thermocouple nearest to the hotplate (held at ~120°C) reached a maximum steady state value of 70°C without a couplant and 80°C with the high temperature couplant; the couplant is required for efficient transmission of sound from the wedge into the pipe. This steady state value is a function of the material heat capacities and conductivities, but also the surface area of the Rexolite.
The results allow the design of a wedge with adequate distance (termed coupling thickness, d) from the specimen and the surface where the probe is attached such that the temperature does not lead to degradation of the active material used in the probe. The results suggest that it will be possible to design a system with a temperature at the surface exposed to the probe equal to or less than ~80°C for a d equal to or greater than 10mm. In addition to the temperature gradients when using high temperature couplants, several other key parameters are important to the ultrasonic system that will need to be established before industrial application, including velocity and attenuation changes in the wedge due to temperature.
The inspection speed using the MAGSCAN (which has a maximum speed of 50mm/sec) can be sufficiently maximised for application at station 2 on a lay barge. Similarly, the data volume from phased array can be balanced between cost and the requirement to archive sufficient detail to re-analyse a particular weld condition at a future date. However, for phased array techniques, the greatest cost in this particular application is likely to be the time required to fully interpret the weld condition to sentence it promptly.
It is estimated that once the data is collected, a trained interpreter will be able to assess the full condition of the weld (for 14" OD pipe) within two to eight minutes, including detection, type identification and sizing. The time will depend on factors such as S/N quality and the number of indications actually present. Accurate evaluation of the analysis time will be subject to extensive qualification effort which was outside the scope of this project.
In recent times the improvements taking place in computing technology and signal processing concepts have allowed the possibility of automating the interpretation of the inspection data. In ultrasonics, work is underway in two branches (namely imaging techniques based on full matrix capture of data 11, 12 and neural network analysis 13, 14) which promise to converge in the near future to make the automated interpretation of data a feasible route, which will require extensive qualifications before industrial implementation.
5.1 Phased array system
The following conclusions draw together the findings in this paper:
- Phased array ultrasonic techniques were successfully developed for inspection of partially completed girth welds in clad pipes. The inspection techniques were verified on 316L clad pipes welded using Duplex 2209 filler. No work was undertaken on nickel alloy welds which will likely prove to be more difficult due to the textured grain structure.
- The high frequency (10MHz) phased array techniques based on the use of longitudinal and shear waves were able to detect the three flaw types studied: porosity, lack of root penetration and incomplete fusion.
- The circumferential sizing accuracy of the PA techniques to detect clustered porosity was within 3mm, confirmed through radiography, sectioning and visual inspection where possible.
- An independent assessment by an EN 473 PAUT Level II operator deemed the lack of root penetration and incomplete fusion sizable in the through-wall direction; however, the through-wall sizing accuracy was not established in this project.
- Metallurgical analysis using the EBSD technique indicates that inspecting through weld-deposited cladding would be more difficult than through metallurgically applied cladding (eg roll bonding) due to the coarse textured grains that develop during the solidification process.
- A study of the probe and wedge system on a surface at an expected operating temperature of 120°C suggests that it would be possible to design a system whose temperature at the surface exposed to the probe is equal to or less than ~80°C
5.2 OMS work
The OMS weld scanning system was evaluated for its capability to detect imperfections or defects in the root welds of pipes as defined by DNV-OS-F101, October 2000. The following conclusions are made:
- The correlation between the phased array data and the OMS data in terms of position and length of flaws appears good.
- The misalignment data measured physically and from the OMS system was 2.5 and 2.61, respectively.
The authors wish to acknowledge the following for funding the work presented in this paper: Air Liquide, Allseas Engineering BV, Applus RTD, BP Exploration Operating Company, Chevron Energy Technology Company, Exxon Mobil Development Company, Japan Steel Works, Petrobras, Pipeline Technique Ltd and Tata Steel.
References and footnote
- M G Silk, A M Stoneham and J A G Temple, ‘The Reliability of Non-destructive Inspection – Assessing the assessment of structures under stress’, Adam Hilger, Bristol, 1987.
- R J Hudgell, ‘Handbook on the ultrasonic examination of austenitic clad steel components’, European Commission Joint Research Centre, 1994.
- R/D Tech, ‘Introduction to phased array ultrasonic technology applications’, R/D Tech Inc, Canada, ISBN 0-9735933-0-X, 2004.
- Zetec Inc.
- Olympus NDT
- Phoenix ISL
- Optical Metrology Services.
- DNV-OS-F101, Submarine Pipeline Systems (2000)
- V Randle, ‘Microtexture determination and its applications’, Institute of Materials, London, 1992.
- J Krautkrämer and H Krautkrämer, ‘Ultrasonic testing of materials’, New York, Springer-Verlag, 1990.
- M Weston, P Mudge, C Davies and A Peyton, ‘Time efficient auto-focussing algorithms for ultrasonic inspection of dual-layered media using Full Matrix Capture’, NDT&E International, Volume 47, pp. 43-50, 2012.
- M Sutcliffe, M Weston, B Dutton, P Charlton and K Donne, ‘Real-time full matrix capture for ultrasonic non-destructive testing with acceleration of post-processing through graphic hardware’, NDT&E International, Volume 51, pp.16-23, 2012.
- F W Margrave, K Rigas, D A Bradley and P Barrowcliffe, ‘The use of neural networks in ultrasonic flaw detection’, Measurement, Volume 25, Issue 2, pp. 143-154, 1999.
- A Masnata and M Sunseri, ‘Neural network classification of flaws detected by ultrasonic means’, NDT&E International, Volume 29, Issue 2, pp. 87-93, 1996.