ISO 9001 2015 Training

Technology Strategy Board
Technology Programme
Guidelines for generating array ultrasonic procedures for the

inspection of dissimilar/austenitic welded components
MATERIALS FOR ENERGY
Phased Array Ultrasonic Inspection of Dissimilar Metal

Joints
TSB Project No: TP11/MFE/6/I/AA058J
Date: 25 January 2011 Ref: TP AA058J-FR

Vs: 01.05
Circulation: Lead organisation: TWI Ltd

Partner 1: EDF Energy
Partner 2: Shell UK
Partner 3: Nuclear Installations Inspectorate (HSE)
Partner 4: University of Birmingham (UoB)
Partner 5: ALBA Ultrasound
Partner 6: Peak NDT
Partner 7: Applied Inspection
Project Monitoring Officer Alan Groves (dstl)
DISSIMILAR 1 TP AA058J-FR
DISSIMILAR Date: 25 January 2011
Ref: TP AA058J-FR
Version: 01.05
Guidelines for generating array ultrasonic procedures for

the inspection of dissimilar/austenitic welded components
General
The following guidelines for designing and implementing ultrasonic inspection of austenitic
dissimilar welds draw together the findings of the Technology Strategy Board funded
research project DISSIMILAR and previously established best practice knowledge.
1. The design and fabrication of a joint containing an austenitic weld, in terms of geometry
and selection of welding processes, should be undertaken with due regard to the
capabilities of ultrasonic inspection techniques.
2. Ultrasonic technique development should be performed on a representative sample

containing the weld to be inspected, fabricated using identical welding procedures.
Ideally, artificially implanted flaws simulating the expected flaws should be used but, if
not possible, then side drilled hole (SDH) targets should be used in the appropriate
positions in the reference weld.
3. The velocity of ultrasound is dependent on propagation direction within an austenitic

weld and the degree of variation depends on the weld. Hence, it is important to get a
handle on this variation in any given weld through basic measurements, for example by
extracting cubic coupons from within the reference weld through sectioning and
establishing the variation in longitudinal (and/or shear) velocities.
Quantifying the weld microstructure using Electron Back Scatter Diffraction (EBSD)
4. To be able to design advanced ultrasonic techniques using arrays, the structure of the
weld must first be quantified. EBSD offers a route (when used in conjunction with
models) with sufficient resolution to capture the beam distortion evident when sound is
propagated through an austenitic weld. This then provides a potential tool for
compensating for or overcoming the microstructural barriers.
5. EBSD is costly (in equipment and time) and the subsequent processing steps required
add further complexity. Specific metallurgical and ultrasonic expertise is required: in
particular the knowledge of the elastic stiffness values (see 47).
6. In order to (1) reduce costs and (2) capture sufficient information to model ultrasonics,
both the scanning resolution and the mapping must be subject to the minimum scatterer
criterion (see Section 3).
7. To make use of the raw EBSD data for application in a semi-analytical model (such as
CIVA) a processing method termed orientation unification must be used, which results in
a map with closed regions containing a limited number of grain orientations and well
defined boundaries. However, use of finite element models may not require this step.
8. An implicit assumption in sampling several weld cross sections is that the microstructure
does not vary significantly along the welding direction. This assumption can be tested
Ref: TP AA058J-FR
Version: 01.05
ultrasonically using a target which runs the length of the weld, monitoring the target at
the beam angles to be used (travelling through the weld), and quantifying changes in
received amplitude and plotted position. The degree of variation will be indicative of any
severe changes in lengthwise uniformity.
Probes
9. The selection of the type of probes will be dictated by the techniques used for inspection.
To achieve the required inspection range, probe frequency must be selected such that
the Rayleigh model can be assumed for the attenuation in the weld, ie the wavelength
must be much larger than the average size of grains in the path of the sound beam (see
also 39).
10. If focusing is to be used, then in the case of single element probes the focal range
should be approximately equal to the required inspection range; more specifically, the
inspection range should generally lie within the 6dB drop zone either side of the focus
point. Additionally, the efficiency of coupling through the probe footprint (ie the surface
through which the sound is transmitted into the component) must be considered. In the
case of arrays, the near zone range due to the aperture must be greater (by at least
10%) than the maximum required inspection range. Note: the aperture is here taken to
be the size of the actual radiating area due to the number of elements chosen for
operation.
11. An iterative approach to the selection of probes is likely to be needed to ensure that the
sound field characteristics at all required inspection regions is sufficient for the task -
detection and/or sizing as appropriate (see 12).
12. Sizing capability is determined by both temporal resolution (ie the pulse length) and the
spatial resolution (ie the beam size at the inspection range). Both of these must be small
enough to achieve the required sizing accuracy. The minimum measurable flaw size will
also depend on these parameters.
13. In the manufacturing specification of the probe, the key parameters to consider for
application to austenitic weld inspection are: (1) pulse length, (2) bandwidth, (3) cross
talk, (4) parasitic echoes, and (5) spatial resolution.
14. Modelling tools are strongly recommended for investigation of the probe capabilities
before manufacture. The use of models (1) reduces likelihood of mistakes in the probe
specifications, (2) allows optimisation of parameters based on given ultrasonic or
geometric constraints, and (3) increases confidence in the resulting techniques, thus
contributing to the technique justification during inspection qualification.
Techniques
15. The technique here refers to the ultrasonic method (covering probes, instrumentation,
physical positioning, encoded / manual scanning, calibration method and setting
sensitivities) that is proposed to satisfy the inspection requirement; see 39 – 50 for
detailed treatment of the component parameters of the ultrasonic technique.
Ref: TP AA058J-FR
Version: 01.05
16. Note that it may not be possible to satisfy the complete inspection requirement with a
single ultrasonic technique and, hence, several different techniques (including other
disciplines such as eddy current) may be required to satisfy the inspection requirement.
17. To be able to design the technique well, expected flaw positions, orientations,
dimensions and characteristics should be known. Techniques should be designed to
meet specific inspection objectives.
18. For inspection of thick section (>25mm) austenitic welds, in particular where the sound is
required to traverse the weld, the use of phased array transmit-receive longitudinal
(TRL) probes is recommended. TRL probes have been shown to provide an
improvement in signal-to-noise (in comparison to conventional dual element probes
which operate on similar principles), along with the versatility to cover a much larger
inspection volume (in comparison to conventional dual element probes which are limited
to a small inspection volume). The achievable inspection volume can be explored
through modelling and probe parameters can then be optimised to achieve the required
capabilities.
19. Two-dimensional (2D) array configurations can be used to achieve full three dimensional
(3D) control of the sound field (note that phased array TRL probes can also achieve 3D
control but are generally much limited in comparison to dedicated 2D arrays). 2D probes
are useful in that they can electronically ‘skew’ the sound beam and hence increase
sensitivity to flaws that do not lie in a plane perpendicular to a non-skewed beam.
20. The 2D probes are particularly attractive for the inspection of flaws that lie close to the
transverse plane to the welding direction, especially in cases where the weld cap
remains in place. Again, to optimise and verify the capability of a particular technique
using 2D probes skewing the beams, modelling must be undertaken to establish
feasibility, optimise parameters and demonstrate capability (through simulation).
21. Techniques based on the full matrix capture (FMC) of data, where each transmit-receive
pair on the array is executed individually, leading to a matrix of data which can then be
post-processed, could also be generated, provided the inspection frequency is well
matched to the material’s attenuation characteristics. Post-processing methods applied
to the matrix of data assume that the sound wave fronts travel uniformly in the medium
and in a geometric straight line (which is not usually the case in austenitic welds). This is
likely to lead to distorted imaging, leading to a similar degradation in inspection quality
as experienced by other ultrasonic techniques.
The adapted delay law (ADL) technique
22. Given that an array probe is able to manipulate the sound field and the austenitic
microstructure influences the sound propagation, the concept of adapted delay laws
(ADLs) is based on the idea that array ultrasonics can be used to overcome the barrier
presented by the microstructure. The method makes use of temporal phasing, ie the act
of delaying the firing time of each element of the array relative to the others; frequency
phasing is not used.
23. ADLs can be generated through two potential routes: (1) modelling and (2)
experimentally. The modelling route requires the microstructure of the weld to be
mapped (using EBSD or other quantification methods), whereas the experimental route
Ref: TP AA058J-FR
Version: 01.05
can be limited by excessive attenuation in the weld metal and hence the inspection
parameters must be well matched to the material attenuation.
24. When generating ADLs through modelling, time reversal concepts are used. The delay
law is built such that the wavelets emanating from each element arrive at the target
region at the same time and phase for constructive interference to take place. First, the
model is used to evaluate the actual time of flight to the target point from each element,
because the actual energy locus is often not a geometric straight line in anisotropic
inhomogeneous materials, such as dissimilar welds. Secondly, the delay law is built
such that the wavelets from all the elements arrive to constructively phase. In essence,
whereas in isotropic media the delay laws can be calculated geometrically, in the
anisotropic media the delay laws have to be generated using the model.
25. Theoretically, ADLs are only applicable to the point for which they were generated
through modelling. Hence to execute inspection of a large volume in a joint, it is
necessary to build a large library of ADLs, which are then used by the array controller
according to the position of the target. The generation of ADLs for this library can be
time consuming (see 26) and hence this adds a significant layer of complexity and cost
compared to standard phased array methods. The cost will be in addition to that for
quantifying the weld microstructure, eg by EBSD.
26. Various modelling packages can be used, subject to validation. At the time of writing, no
known models have been specifically validated for the purpose of generating ADLs. A
range of models exist (eg the semi-analytical CIVA and the finite element codes of
ABAQUS and PZ Flex) that are able to propagate sound in complex media but some
models are better than others in terms of computation times (for similar resources).
27. Experimentally, ADLs can also be generated subject to the caveat in 28. To achieve this,
firstly targets (eg SDHs) must be introduced at the inspection sites. Then each element
of the array can be fired individually and the arrival time of the echo from the target can
be recorded. Then the ADLs can be generated as before by considering that all the
wavelets must arrive at the target region concurrently. The experimental data collection
method is a subset of full matrix capture of data (see 21) because only the pulse-echo
signal from each element is required.
28. An assumption in deriving the ADL experimentally is that the echo from the target can be
clearly identified and is distinct from other echoes. Since there is likely to be significant
back scattered noise, identification of the relevant echo will not always be
straightforward. Additionally, the size of elements in typical array probes may not be well
suited to penetrate to any significant distance within the weld and hence no signal may
be received at the required ranges. In general, for thick section austenitic welds, the
experimental route for generating ADLs may not be physically well conditioned and
selection of this route must be subject to stringent qualification evidence.
Personnel and training
29. Personnel who undertake the inspection of the components must be qualified to at least
Level 2 through a certification scheme governed by EN 473. Additionally, the operators
should undergo component-specific training (through the use of representative mock-ups
– see 31) using the probes, techniques and procedures specified for use in the
inspection. The component-specific training must be suitably monitored and the
Ref: TP AA058J-FR
Version: 01.05
performance of the personnel must be assessed to ensure they meet minimum

requirements.
30. The minimum requirements required of the inspection personnel must be developed by a
Level 3 holder (compliant with EN 473), subject to a review of the inspection
requirements as stated in the procedures. The assessment of the inspection personnel
must be witnessed by the Level 3 and approval authorised independently.
31. Representative specimens (using identical materials, welding procedures and ideally of
identical geometries and sizes) must be generated containing representative flaws.
These mock-ups must be used for the qualification of techniques and for the training of
the inspection personnel.
32. The inspection personnel must be made aware of the issues involved in the inspection of
inhomogeneous coarse grained anisotropic materials through relevant literature and
practical demonstrations. Their performance in the component-specific training (see 29)
must be assessed in terms of not just the detection of flaws but also the errors in
positioning, sizing and characterisation (where applicable).
33. The personnel who develop the techniques (including the selection of conventional or
specialist ultrasonic technology) must be overseen by a Level 3. Tasks involving EBSD
and the development of ADLs may involve personnel who are not qualified in ultrasonic
inspection. Hence, the aim of the guidance provided by the Level 3 is to ensure that the
decisions taken during the development process are considered within their proper
context.
Validation and qualification
34. The techniques and procedures developed should ideally be subject to stringent
qualification. The qualification methods should be subject to standards specified by
either the European Network for Inspection Qualification (ENIQ) (see 35) or the
American Society of Mechanical Engineers (ASME) Boiler and Pressure Vessel Code
Section XI or Section V, as appropriate (see 36). Open trials for procedure qualification
may be performed on the mock-ups generated for the specific training (see 31).
However, it may be appropriate to also involve other qualification specimens, eg if blind
trials are required (see also 35 and 36).
35. The ENIQ route to qualification requires the definition of inspection objectives. The
concepts being proposed in the inspection should be subject to technical justification,
which may involve the use of simulations (subject to the caveats in 37) and experiments.
The technical justification is designed to provide reasonable evidence that the
techniques being proposed will meet the inspection objectives. ASME V Article 14
contains similar requirements.
36. The ASME XI route to qualification calls for the use of performance demonstration test
pieces containing representative flaws. The component should be representative (as in
31) and the flaws used should be appropriately positioned in regions of the weld where
damage could be critical to the operation of the component.
37. Models used for inspection simulations should be independently validated. Scenarios
used in the validation efforts must correspond to the inspection scenario as closely as
Ref: TP AA058J-FR
Version: 01.05
possible but need not be identical; the requirement is to establish a scientific judgment of
relevance between the simulation to be performed and the scenarios used in the
validation. The validation evidence should ideally include comparisons with experiment,
but may also include comparisons with previously validated models. The limits to the
validity of the models should be clearly established and the accuracy of the models
should be quantified and clearly documented.
38. The use of EBSD and ADLs are considered special cases and their use will need to be
subject to stringent evidence also. These technologies for use in ultrasonic inspections
require special care as they lead to significant complications if implemented incorrectly.
Key parameters for technique design
39. Selecting the sound frequency is dependent on two primary considerations. Firstly, the
frequency must be well conditioned with respect to grain size (see 9), as scattering of
the sound waves at the grain boundaries is the dominant component of attenuation
(absorption being the other). Secondly, the resolution of the technique to detect small
flaws is dependent on the wavelength (hence frequency) and so the frequency must be
high enough to be sensitive to the minimum flaw size that must be detected.
40. Considerable past evidence has shown that the longitudinal wave mode is often better
suited to inspection of coarse grained anisotropic materials than the shear mode (for the
same inspection frequency), as the wavelength is larger in comparison to the shear
mode since the sound velocity is faster. Additionally, the angular difference between
group velocity and phase velocity due to anisotropy are larger for the vertically polarised
shear wave than for the longitudinal wave or the horizontally polarised shear wave.
Where the coarse grained volume is not significant (eg the root) it may be possible to
use the shear wave modes; in particular, the horizontally polarised shear wave has been
shown to be a feasible mode for inspection.
41. The beam angle(s), measured with respect to the vertical at the point of entry – ie the
index point – into the component, should be chosen to firstly detect the flaws and
secondly to size them (assuming sizing is required). The reflectivity (including signals
generated from diffraction effects) of flaws is dependent on the incident beam angle and
on many other geometric conditions, such as whether the flaw is void-like, crack-like and
whether it is close to material boundaries such as a component surface etc. In general,
the beam angle should be selected to maximise the amplitude of the reflected echo.
Models can be used to aid in technique design to select beam angle(s) to achieve the
goals of detection and/or sizing.
42. The scanning extent should be sufficient to cover 100% of the volume (weld or
otherwise) where flaws that threaten the integrity of the component or structure could
develop either during service or post-fabrication. The required inspection volume should
be covered by either the physical movement of the transducers using scanners or
electronic manipulation of the sound field, or a combination of both.
43. The approach to the calibration of the system (ie the time base) will depend on the
severity of distortion induced by the medium. In a medium where changes in velocity due
to the anisotropy are significant, average values of the sound velocity should be used for
plotting and data presentation. This route will lead to errors in positioning echoes and the
Ref: TP AA058J-FR
Version: 01.05
resulting inaccuracies in sizing should be quantified during any qualification of the sizing
techniques (using representative specimens).
44. Sensitivity for the inspection should be set using calibration targets (typically SDHs)
placed in the representative specimen in locations where the flaws are expected. The
diameter of the SDHs will be similar to the minimum size of flaw requiring detection, and
will usually lie in the range 2 to 5mm. A specimen containing a representative weld shall
be used so that the attenuation effects of the weld material are adequately captured in
the sensitivity setting for the inspection.
45. Techniques may make use of wedges to generate beam angles inside the component
using the law of refraction. The wedge material may be solid (eg Perspex or Rexolite) or
liquid (eg water). A water wedge has the advantage of tolerating inspection surface
roughness (see 46) better than a solid shaped wedge. Solid wedges also require the use
of couplant material (eg Ultragel II) between the wedge and the inspection surface. The
distribution (thickness) of couplant should be consistently maintained during calibration,
sensitivity setting and the actual inspection.
46. The surface roughness shall be less than or equal to 6µm Ra; where this level is not
achieved, it is recommended that the surface is adequately prepared before inspection
takes place. The use of the longitudinal wave mode (see 40) may lead to mode
conversions on reflection and refraction at boundaries, which could complicate the
inspection. Hence, techniques using the longitudinal wave mode may require the weld
cap to be ground flush to allow access to 100% of the weld volume using direct
incidence on flaws, ie without prior reflections (skipping) at component boundaries. If,
however, the weld cap cannot be removed, then techniques that take due account of
mode conversion effects and/or polarised shear wave modes may need to be
considered.
47. The elastic stiffness constants are here considered to be a key parameter for effective
technique design only when utilising the EBSD method to quantify the weld and
generate ADLs. The austenitic weld demonstrates anisotropic properties which derive
from the anisotropy of the face centred cubic (FCC) unit crystal, whose elastic stiffness
constants can be evaluated using a single crystal of the alloy and measurements of
ultrasonic velocity. This approach can incur significant costs. Alternative approaches to
the evaluation of elastic stiffness constants include the use of values from the literature,
where these have already been evaluated for alloys of similar composition.
Ref: TP AA058J-FR
Version: 01.05
Contents
Executive summary
Background
Objective
The industrial case
Ultrasonic phased array technology
Contributing authors and their affiliations
1 INTRODUCTION 15
2 WELDED COMPONENT 17
2.1 JOINTS WITH AUSTENITIC DEPOSITS 17

2.2 JOINT CONFIGURATION OF THE DISSIMILAR PROJECT 17
3 MICROSTRUCTURAL ANALYSIS 25
3.1 INTRODUCTORY COMMENTS 25

3.2 EBSD APPROACH TO QUANTIFYING THE WELD 27
3.3 OPTIMISATION OF EBSD SCANNING PARAMETERS 29
3.3.1 THE MINIMUM SCATTERER CRITERION 30
3.3.2 EMPIRICAL STUDY OF SCANNING RESOLUTIONS 34
3.4 PROCESSING THE EBSD DATA 34
3.5 ANALYSIS OF THE MICROSTRUCTURE AND TEXTURE OF THE AUSTENITIC WELD 36
3.6 LENGTHWISE UNIFORMITY 39
3.6.1 ANALYSIS OF LENGTHWISE UNIFORMITY USING EBSD 39
3.6.2 ANALYSIS OF LENGTHWISE UNIFORMITY USING ULTRASONICS 43
3.6.3 SUMMARY OF LENGTHWISE UNIFORMITY ANALYSIS 49
3.7 EVALUATION OF THE STIFFNESS CONSTANTS 49
3.8 SUMMARY OF EBSD ANALYSIS 50
3.9 FINANCIAL COSTS OF IMPLEMENTING EBSD 50
4 PROBES 51

4.2 REQUIREMENTS 51
4.3 SPECIFICATION APPROACH 52
4.4 SPECIFICATION OF THE TRL-1 ARRAY 52
4.4.1 FIRST LEVEL SPECIFICATION (CIVA AND SIMULUS) 52
4.4.2 SECOND LEVEL SPECIFICATION (PZ FLEX) 74
4.4.3 THE TRL-1 PROTOTYPE FOR IMMERSION COUPLED INSPECTION 77
4.5 2D-1 ARRAY 81
4.6 ADVANCED ARRAY CONFIGURATIONS FOR FUTURE EXPLORATION 82
4.7 SPECIFICATION FLOW CHART AND STANDARDISATION OF PROBES 83
5 INSTRUMENTATION 85
Ref: TP AA058J-FR
Version: 01.05
6 TECHNIQUES 86

6.2 ADAPTED DELAY LAWS (ADL) TECHNIQUE 87
6.2.1 INPUTTING THE QUANTIFIED WELD TO THE MODEL 88
6.2.2 GENERATING THE ADL 89
6.2.3 EXAMPLE OF INSPECTING FLAW 3 USING ADL 89
6.3 BASELINE TECHNIQUES 92
6.3.1 MANUAL CONVENTIONAL TECHNIQUE 93
6.3.2 AUTOMATED (ENCODED) CONVENTIONAL TECHNIQUE 93
6.3.3 LINEAR PHASED ARRAY TECHNIQUE 93
6.3.4 TRL-1 PHASED ARRAY TECHNIQUE 94
6.3.5 2D-1 PHASED ARRAY TECHNIQUE 94
7 PERFORMANCE 95

7.2 BASELINE INSPECTIONS 95
7.2.1 FLAW 2 96
7.2.2 FLAW 3 103
7.2.3 FLAW 4 108
7.2.4 FLAW 5 112
7.2.5 FLAW 6 115
7.2.6 FLAW 7 118
7.2.7 FLAW 8 121
7.2.8 GENERAL DISCUSSION 123
7.3 ADAPTED DELAY LAWS (ADL) TECHNIQUE 127
7.3.1 SIMULATED 127
7.3.2 EXPERIMENTAL 144
7.4 MODEL VALIDATION 150
8 CONCLUSIONS 157
9 RECOMMENDATIONS 158
10 FUTURE DIRECTIONS 159
10.1 ARRAY PROBE CAPABILITIES 159

10.2 INSPECTION TECHNIQUES BASED ON FULL MATRIX CAPTURE OF DATA 162
10.3 FINITE ELEMENT MODELLING PACKAGES 166
11 REFERENCES 170
Appendix A: Ultrasonic inspection of an austenitic weld - a case study

Appendix B: Specification of the TRL-1 array
Appendix C: Specification of the 2D-1 array
Appendix D: Specification of the MicroPulse 5PA
Appendix E: Automated baseline inspection report (British Energy)
Ref: TP AA058J-FR
Version: 01.05
Executive summary
Background
This document was generated in the DISSIMILAR project (TSB project no.
TP11/MFE/6/I/AA058J) and aims to outline the method developed in the project for ultrasonic
inspection of textured coarse grained austenitic welds. This document is not a generic
technical justification for inspection of these welds through the use of phased arrays; the
limited data generated in the DISSIMILAR project is used to illustrate and propose best
practice routes available for the inspection of such joints. In effect, this document outlines a
set of options available to the ultrasonic inspector when faced with inspecting what is widely
known to be a challenging class of fusion joints between metals.
This document will make reference to several earlier documents generated in various
different organisations and forums. The use of evidence and recommendations from the
reference documents will be selective and this use does not implicitly support all
aspects/claims of those documents.
The DISSIMILAR project (www.dissimilarweld.co.uk) was initiated in July of 2008 and

completed in December of 2010. The project Consortium was composed of a total of eight
Partners. Three of the Partners - British Energy Generation Ltd, Shell UK and the Nuclear
Installations Inspectorate of the Health & Safety Executive - formed the Project Steering
Committee under whose guidance this document has been finalised. The bulk of the
research and development focused on the inspection of a specific component and was
undertaken by the Metallurgy and Materials department of the University of Birmingham and
TWI Ltd. Development of ultrasonic transducers (probes) was undertaken by Alba
Ultrasound, instrumentation was developed by Peak NDT and implementation of baseline
scanning and performance trials was done by Applied Inspection.
The aim of the project was to:
1. Position and size flaws/accurately, compensating for the distortive effects of the
anisotropic, inhomogeneous austenitic/ferritic weld.
2. Improve the overall inspection quality through the use of phased array technology
such that the inspections provide vastly better signal-to-noise quality (compared to
present probes), are faster (by orders of magnitude) and provide accurate,
quantifiable and digitally recordable data.
The project was focused on a class of joints termed ‘dissimilar’ or ‘transition’ welds, which
generally refer to the joint between two different types of materials. In the context of this
project, the joint is between parent ferritic and parent stainless steels. The weld metal
between the two parents is in the austenitic condition. It is widely known that inspection of
austenitic welds is difficult in comparison to equivalent ferritic joints, but those involving the
use of buttering layers have proven to be even more difficult to inspect.
Objective
The aim of this document is to provide guidance, based on the experience of the
DISSIMILAR project, for the inspection of austenitic / dissimilar welds and disseminate the
generated data, results, methods and findings.
Ref: TP AA058J-FR
Version: 01.05
The industrial case
The DISSIMILAR project was focused on the energy sectors in the UK to meet the low
carbon vision of the future outlined by the Government of the United Kingdom. The ability to
design the new components for a high stress, high temperature environment through
Engineering Critical Assessment (ECA) methods depends directly on the ability of the
inspection techniques to detect and accurately size the flaws.
In the thermal energy sectors (ie oil & gas and nuclear) the use of dissimilar welds in critical
joints has initiated the drive to improve the capabilities of the non-destructive testing (NDT)
methods, in particular the application of ultrasonics. Line pipes used to transport the products
extracted from within the Earth increasingly need to be clad with a corrosion resistant alloy
(CRA) to survive chemical attack by the products. The joint between these clad pipe sections
(the girth welds) often require the use of weld fillers such as Inconel 625, 316 and super
duplex 2209. Depending on the weld metal volume, bevel geometry, thickness and the filler
grade, the inspection of clad line pipe girth welds vary in difficulty from possible to severely
limited possibilities.
Those critical dissimilar joints which make use of buttering layers between the carbon
(ferritic) and the weld are known to be even more difficult to inspect as the ultrasound finds it
difficult to traverse the buttering layer. In the oil & gas industry a key example of such a weld
can be found on the sea bed where flow lines are collected at a junction termed the subsea
hub before being despatched upwards to storage facilities. The use of dissimilar joints
involving buttering has been known to cause problems (TWI Member’s Report 962/2010) at
the buttering layer with the need to detect very small flaws induced by hydrogen
embrittlement. The detection and characterisation of such small flaws (sub one millimetre) is
a significant challenge for ultrasonic NDT and the development work undertaken as part of
the DISSIMILAR project goes some way to address the issues.
Within the nuclear industry there are a number of joints close to the reactor pressure vessel
(RPV) and within the primary cooling circuit that are classed to be critical for containment of
radiation. The joint which connects the CRA clad reactor pressure vessel to the stainless
steel pipe work is a classic thick sectioned dissimilar joint including buttering. The specimen
generated for the developmental work in the DISSIMILAR weld made use of the welding
consumables and procedure to replicate the surge nozzle safe end transition weld. However,
the DISSIMILAR specimen was twice as thick as the joint system in actual use so that the
effects of the weld on the sound propagation could be better explored.
In addition to dealing with the configurations that are currently in service, the project was also
aimed at pushing the boundaries of ultrasonic capabilities for application to inspecting the
future configurations in the European Pressurised Reactor and the AP1000 designs, as well
as for the ITER fusion reactor. The need to improve efficiency requires operation at higher
temperatures which places increased mechanical stresses on the joints, which in turn
requires the evaluation of the joint integrity to be improved. The presence of any cracks (both
post-manufacture and those that may develop in-service) will be deemed a serious threat to
the integrity of the nuclear components by the regulatory bodies, given the increased
potential for its growth leading to catastrophic failure in the increased stress environment.
Hence the aim of the ultrasonic inspection programme will be to first detect the presence of
(primarily) crack-like flaws and then accurately characterise them (ie evaluate dimensions
and orientations), such that their threat to the integrity of the components can be evaluated
with greater confidence.
Ref: TP AA058J-FR
Version: 01.05
Ultrasonic phased array technology
Ultrasonic technology has been widely used from the earliest days for the inspection of
austenitic materials. A key advantage of ultrasonics, in comparison to radiography, is the
ability to better characterise crack-like flaws as the interpretation is based on signal
amplitudes, orientations and distances (based on times of flight); whereas in radiography the
interpretation is based on absorption (which can be severe) and the ability to evaluate crack-
like flaws is limited. In comparison to techniques such as eddy currents or magnetic particle
inspection, ultrasonic techniques are not limited to flaws which exist at or very near the
surface. However, it is important to note that all techniques have their place and, for
example, eddy currents are far more suited to the detection of surface breaking cracks (eg
toe cracks) than ultrasonics and so should be used to complement wherever possible.
Phased array ultrasonic technology is a sophistication of the conventional ultrasonic

technology and the key difference is in the probes. Conventional probes are designed to
generate a sound beam along a certain angle when placed on the component. The sound
beam may or may not be designed to concentrate the acoustic pressure at a certain point
(termed focusing). Using array probes, however, the sound can be ‘steered’ to a range of
angles within the component and can also be ‘focused’ to concentrate the energy to a
desired region. Hence, firstly, the array probe offers greater versatility in that one probe can
replace several conventional probes. Array probes can also be used in other configurations,
including electronic scanning and full matrix capture of the data, which allow for the
possibility to achieve a more sophisticated ultrasonic inspection in comparison to the use of
conventional single beam angle probes.
The key drivers for adoption of the technology for inspection of critical components derive
from demonstrated ability to reduce inspection times, hence direct costs and those due to
lost operation time. Reducing the inspection time has an additional advantage in the nuclear
sector as radiation dosage legislation in Europe is getting stringent and there is a drive to
minimise human exposure to radiation in all activities. Mechanically, the ability to build
systems which need not carry a large number of probes reduces the chances of creating
debris within containment areas.
Phased array technology is, however, still expensive to implement. The cost lies in procuring
the probes (which are orders of magnitude more expensive than conventional probes) and
the instrumentation (termed array controllers) which are also several orders more expensive
than standard conventional probe instrumentation. Additionally, there are significant costs in
training personnel to use the technology, especially in the field of data interpretation. Very
often the data collectors and the data interpreters are different personnel and there is a
requirement to establish the working mechanisms.
Finally, the development of inspection procedures using phased array techniques is critical
for the inspection to be successful. Procedure development incorporates the selection of the
array probe ensuring that the sound field characteristics are sufficient and unwanted energy
distributions (termed side energy lobes) do not degrade inspection quality. Very often mode
conversion phenomenon can adversely affect the interpretation if not clearly understood and
accounted for in the procedure. The use of modelling tools to evaluate the inspection
capabilities is much more important when developing phased array techniques in comparison
to conventional ultrasonic techniques. The course of the DISSIMILAR project, which follows,
is an illustration of the stages required to implement a phased array solution.
Ref: TP AA058J-FR
Version: 01.05
Contributing authors and their affiliations
Bartlett , Alan Applied Inspection

Carpentier, Capucine TWI Ltd
Druce, Stephen Nuclear Installations Inspectorate (HSE)*
Miller, Malcolm Shell UK
Nageswaran, Channa TWI Ltd
Nicholson, Ray Nuclear Installations Inspectorate (HSE)*
Pearce , John EDF Energy
Tse, Yau Yau University of Birmingham
Tweedie, Andrew Alba Ultrasound
Weston, Miles TWI Ltd
Whittle , Alison Peak NDT
The authors would like to gratefully acknowledge the reviews of the guidelines and editing
provided by Charles Schneider of TWI to complete this document.
* The Nuclear Installations Inspectorate (NII) of the Health and Safety Executive (HSE)
supports this work to promote development of new and or improved NDT capability. The
guidance that follows from the project is technical judgements coming from the participants.
HSE NII is not prescriptive in how licensees present a safety justification and therefore would
not wish to imply endorsement of the guidelines from a regulatory standpoint.
Ref: TP AA058J-FR
Version: 01.05
1 Introduction
Ultrasonic inspection techniques are widely used for the non-destructive testing (NDT) of
welds composed of austenitic material. The weld is defined as a fusion joint of two materials
which is created by the deposition of a pool of a melted alloy which subsequently solidifies to
form a metallic bond between the two materials being joined. The use of austenitic materials
is dictated by the mechanical and environmental (high temperature / corrosive) conditions of
nuclear power plants and products of the oil & gas (O&G) industries. However, the
propagation of the ultrasonic wave through the austenitic material is complicated by its
inhomogeneous and anisotropic nature. The primary obstacles to the uniform propagation of
the sound wave front are the large (coarse) grains that develop during solidification of the
weld and the differential texture (anisotropy) that exists across the grain boundaries.
‘Austenitic’ refers to a phase state of the element iron in which it takes the face-centred cubic
(FCC) unit crystal atomic structure. Iron exists in its austenitic (or γ-phase) at a range of
temperatures in the region of 1000°C; the addition of the key alloying element nickel retains
this phase on cooling to room temperatures. The thermal cooling rates and channels give
rise to what is termed dendritic epitaxial grain growth over successive weld deposits leading
to the coarse elongated grain structure characteristic of austenitic welds. The solidified weld
with the FCC structure exhibits strong anisotropy, ie it exhibits differing elastic properties
depending on the direction in which it is stressed.
An ultrasonic wave propagates by the mechanical vibration of the atoms in the medium in its
path such that the energy exists in a kinetic form. The mechanical motion of the atoms is
dictated by the elastic stiffness between them, ie the elastic modulus. In the austenitic
material of the weld, the propagating sound wave experiences changing elastic modulus
along its path, leading to the three well known consequences for their ultrasonic inspection
(in comparison to inspection of typical ferritic steel materials):
1. Increased backscattered energy, primarily from the grain boundaries which exhibit
significant differences in texture across them; this energy increases the ‘noise’ in the
received ultrasonic signals.
2. Increased attenuation of the propagating sound wave, limiting the range of inspection.
Attenuation is composed of absorption of the energy in the wave by the material and
losses due to scattering. In austenitic material, it is the increased backscatter of
energy which is the primary contributor to increased attenuation.
3. Increased distortion of the sound wave, including what is termed ‘beam skewing’. In
essence, whereas in isotropic ferritic material the wave front would propagate
uniformly, in the anisotropic austenitic material this is no longer the case.
In summary, the inspection of austenitic welds, in particular thick section welds with
thicknesses in excess of 25mm can be difficult. The welding processes and conditions have
a strong influence on the microstructure that develops in the weld upon solidification, and the
microstructure then dictates the values of key parameters of the ultrasonic system. Appendix
A contains a case study which illustrates the phenomenon described above.
It is noted as a key recommendation that, ideally, a holistic design of an austenitic joint -

selection of geometries and welding processes - should take into account the ultrasonic
inspection capabilities and this document aims to highlight those key aspects that will be of
interest to inspection personnel, joint designers and welding fabricators.
Ref: TP AA058J-FR
Version: 01.05
A good treatise on the subject of ultrasonic inspection of austenitic welds is given in the
Handbook by the International Institute of Welding [1], covering aspects of the material,
issues, techniques and guidelines. In many ways this document complements the Handbook
[1] by illustrating the design of an inspection technique for an austenitic weld. However, since
this document is based on the specific case tackled in the DISSIMILAR project, the reader is
referred to the Handbook [1] for a much wider discussion of the topic.
The primary aim of this document is to provide guidelines. The essential components of the
ultrasonic inspection system are covered in their own sections and the recommendations are
collated in Section 9. The secondary aim of this document to disseminate the data, results,
methods and findings is also done through these sections.
The particular welded component considered in the DISSIMILAR project is described in

Section 2, followed by its microstructural analysis in Section 3 using Electron Back Scatter
Diffraction (EBSD). The use of EBSD was pioneered for application to the task of ultrasonic
inspection development in the DISSIMILAR project and it now allows the possibility to
quantify the complex structure of an austenitic weld.
Sections 4 and 5 consider the physical hardware requirements for the inspection, ie the
probes and instrumentation, which are borne out of the microstructural analysis of the welded
component. The necessary performance characteristics (bandwidth and sensitivity) required
of the piezocomposite array probes is discussed in Section 4. The electronic capabilities
required of the array controller are outlined in Section 5.
The techniques are then described and developed in Section 6. Several existing techniques
were used in the DISSIMILAR project (as baselines) and an original technique (termed
adapted delay laws) based on time reversal concepts was developed as part of the project.
This was made possible by (1) ability (through EBSD) to quantify a weld, (2) through a model
(called CIVA, developed by Commissariat à l’Énergie Atomique) to simulate the propagation
of sound through that weld and (3) by array ultrasonic probes, which all together allow the
possibility to manipulate the sound propagation to overcome the distortion induced by the
weld.
Section 7 contains the experimental data generated in the project and contains an analysis of
the performance of all the baseline techniques individually and in comparison to each other.
Section 7 also discusses the performance of the adapted delay law technique with regard to
its ability to deal with the distortion and improve the inspection quality.
Finally, Section 10 provides a discussion of future directions for efforts in this topic,
consolidating the work done in the DISSIMILAR project to outline areas where further work is
required.
This document is envisaged as a full account of the course of the DISSIMILAR project and
aims to disseminate the knowledge to a wider audience to recreate, critically analyse and
improve upon. The inspection of austenitic welds remains a challenge but, undoubtedly, the
ever continuing improvements in the various technological fields considered in this document
will eventually overcome the many obstacles.
It is hoped that, the guidelines and findings presented in this document will aid the future
design of austenitic welded joints and improve confidence in their ultrasonic inspection.
Ref: TP AA058J-FR
Version: 01.05
2 Welded component
2.1 Joints with austenitic deposits
There are several key joint configurations used in both the nuclear and O&G industries that
contain weld deposits which are in the austenitic state at operating temperatures. These
include:
1. The safe-end welds, which connect the nuclear pressure vessel to its primary cooling
circuit.
2. Subsea joints on hubs/manifolds, the use of which is rapidly increasing with the need
to access oil/gas fields in deep waters.
3. Corrosion resistant alloy (CRA) clad flow lines which require the use of austenitic
weld fillers in their girth welds due to metallurgical reasons.
There will be numerous other joints which contain austenitic weld deposits and hence many
of the issues illustrated in Appendix A. Additionally, the content of this document will be of
interest to the ultrasonic inspection of castings which develop large epitaxial grains and
electroslag welds for similar reasons.
The term dissimilar metal welds (DMWs) is in common use within both the nuclear and O&G
industries and refers to cases where the joint is between a low-alloy ferritic steel and an
austenitic stainless steel (such as the specimen studied in Appendix A). Both the safe-end
and subsea joints fall within this class of joints. A particular feature of some DMWs is the use
of buttering between the ferritic component and the weld deposit which is deposited before
the weld and often from a different welding position. It is known that the presence of buttering
layers can lead to further inspection challenges in addition to those presented by
homogenous austenitic welds.
Clad flow lines (line pipes) are not considered to be in the category of DMWs but have very
specific inspection requirements which are complicated by the difficulties of inspecting
austenitic material. In this document the specific case of line pipes is not discussed but the
reader will be able to gain some insight into designing their ultrasonic inspection effectively.
2.2 Joint configuration of the DISSIMILAR project
The joint elected for use in the DISSIMILAR project is representative of a DMW used in the
nozzle safe-end transition. However, the specimen fabricated for investigation in the project
was thicker (around 85mm) than the dissimilar weld in-service, which is only 41mm thick (see
specimen in Appendix A). The welding procedures aimed to accurately recreate the original
fabrication and thus simulate the conditions present in the actual weld which is presently in-
service. The specimen was manufactured as a plate with artificially implanted flaws for the
purposes of:
Simulating typical weld flaws

Exploring the capabilities and limitations of ultrasonic inspection
Studying the microstructure of the weld and its properties (Section 3)
Developing techniques to inspect the joint
Evaluating the performance of the techniques (Section 7)
Once manufactured the specimen was halved into what is termed Half-A and Half-B for ease
of handling and a reference piece near the middle of the specimen was extracted.
Ref: TP AA058J-FR
Version: 01.05
The methods used to implant flaws were developed by Sonaspection (UK) and will not be
discussed in this document. It is recommended that artificial flaws of typical weld flaws such
as lack of fusion, root flaws and rough faceted cracks be implanted at expected positions in a
representative weld for technique development. At the very least, technique development
should make use of a reference piece with side drilled hole (SDH) targets in a representative
weld fabricated using the actual welding procedure used for the fabrication of the weld to be
inspected. This is necessary as the effects of the weld microstructure (as illustrated in
Appendix A) can be quite severe and are very sensitive to welding procedures.
Figure 1 shows the weld preparation of the DISSIMILAR specimen. The joint is between a
50D carbon steel plate and a 316L stainless steel plate. Deposition of cladding, buttering and
weld was all done to ASME III guidelines. The V-prep butt weld had a 30° included angle and
was deposited in the flat 1G position.
Figure 1 Weld preparation of the DISSIMILAR specimen.
The deposition procedures follow and are representative of fabrication techniques used for
typical DMWs used in both the nuclear and O&G industries.
Cladding procedure
Process GTAW.
Horizontal 2G (to ASME III).
Consumables:
Sandvik 24.13 LHF (Type 309L) 1.2mm dia. wire
Sandvik 19.9L (Type 308L) 1.2mm dia. wire
Pre-heat: 150˚C min, interpass 150 min - 232˚C max.
Heat treatment: 593-621˚C.
Ref: TP AA058J-FR
Version: 01.05
Figure 2 Details of the procedure for cladding the 50D plate.
Buttering procedure
Process SMAW.
Flat 1G (to ASME III).
Consumables:
Inconel 182 electrodes (E Ni.Cr.Fe.3).
Pre-heat: 150˚C min, interpass 150 min - 232˚C max.
Heat treatment: 593-621˚C.
Figure 3 Details of the procedure for buttering the 50D bevel face.
Ref: TP AA058J-FR
Version: 01.05
Welding procedure
Process GTAW and SMAW.

Flat 1G (to ASME III).
Consumables:
Inconel 82, 2.4 dia. wire (ER Ni.Cr.3).
Inconel 182 electrodes (E Ni.Cr.Fe.3).
Pre-heat: 5˚C min, interpass 232˚ max.
Heat treatment: None.
Cap and root will be dressed flush.
Figure 4 Details of the procedure for filling the weld.
A total of eight flaws were artificially introduced into the specimen. The lengths and positions
of all eight flaws when looked at from above the weld are shown in Figure 5. Flaw 1 is an
electrical discharge machined (EDM) notch designed for the study of the microstructure
ultrasonically (Section 3). Flaws 2, 3, 4, 5 and 6 are lack of fusion (LOF) type flaws on the
different fusion boundaries. Flaw 7 was designed to be a rough (multi-faceted) crack along
the centreline of the weld and Flaw 8 is a surface breaking vertical (EDM) flaw at the root.
It is important to note that the insertion of the flaws will be affected by the welding, and their
presence can also affect the welding and the resultant microstructure as they would modify
the heat extraction flows.
Each of the flaws above was introduced with the aim to study the inspection capabilities.
Figures 6 to 13 provide through thickness positions, sizes and orientations of all eight flaws.
Note that Figures 6 to 13 show the intended position of the flaws; the actual position of the
flaws (determined through sectioning) is presented in Section 7 as part of the performance
analysis.
Figure 14 shows the as-received specimen post-fabrication.
The key parameters of interest to ultrasonic inspection of the component are (1) thickness,
(2) materials and (3) bevel geometry. These parameters have to be determined primarily
through the mechanical stress conditions the component is required to withstand and
metallurgical considerations. However, if sufficient thought is given to the inspection
requirements as well during the early design stage of the joint then the subsequent
inspection techniques and procedures are likely to be much better configured to assess the
condition of the joint throughout its lifetime.
Ref: TP AA058J-FR
Version: 01.05
Figure 5 Specimen dimensions and flaw positions; Flaw 1 was as long as the specimen
whereas all other flaws were 25mm long.
Figure 6 Flaw 1, EDM slot in the stainless steel parent material.
Ref: TP AA058J-FR
Version: 01.05
Figure 7 Flaw 2, smooth LOF on the external buttering fusion face (carbon steel/buttering).
Figure 8 Flaw 3, smooth LOF on the internal buttering fusion face (buttering/weld).
Figure 9 Flaw 4, smooth LOF on the weld/stainless steel fusion face.
Figure 10 Flaw 5, smooth LOF on the external buttering fusion face (carbon steel/buttering).
Ref: TP AA058J-FR
Version: 01.05
Figure 11 Flaw 6, smooth LOF on the weld/stainless steel fusion face.
Figure 12 Flaw 7, rough fatigue crack like flaw along the weld centre line.
Figure 13 Flaw 8, smooth lack of root fusion.
Ref: TP AA058J-FR
Version: 01.05
Figure 14 Image of the as-received DISSIMILAR test specimen.
Ref: TP AA058J-FR
Version: 01.05
3 Microstructural analysis
3.1 Introductory comments
Ultrasonic velocity (both shear and longitudinal) are functions of the material properties
elasticity (bulk / shear), density and Poisson’s ratio [2]. Changes in velocity along different
directions imply changes in material properties along those directions. Assuming that bulk
density and Poisson’s ratio remain unchanged, then the changes in velocity are due to the
changes in elasticity. Figure 15 shows a 13mm thick cross section of a K-prep DMW along
with the dimensions.
Figure 15 A 13mm thick section of a K-prep weld used in a nuclear power plant.
Considered separately the different components of the DMW will exhibit different longitudinal
velocities; typical measured values are given below:
Ferritic nozzle - 5892m/s

Inconel buttering and weld metal - 5890m/w
Austenitic safe-end stainless steel - 5740m/s
The weld shown in Figure 15 was then divided into 13 accurately machined cubes (ie the
cubes are weld metal), as illustrated in Figure 16. The longitudinal velocity of along the three
possible directions through the cube was then measured in immersion using the second and
third back wall echoes.
The mean of the velocities measured through all sides and through all the cubes was
evaluated to be 5910m/s, which is consistent with the longitudinal velocity in steel. However,
Figure 17 shows the variation in velocity through each side of all the cubes expressed as a
difference from this mean velocity. The extent of the variation is dependent on many factors
and will be different between different welds. However, the existence of the variation is well
known: for instance lesser variation from 5500 - 6100m/s was reported for another austenitic
weld [2]. Conventional ultrasonic techniques assume that the velocity of the sound, once
introduced into the material, remains constant. Out of this assumption the tasks of plotting
the position of echoes (indications) can be effectively carried out. Clearly, in an austenitic
Ref: TP AA058J-FR
Version: 01.05
weld, the isotropic assumption of constant velocity is no longer valid and will lead to errors in
geometrical plotting.
Figure 16 Division of weld into cubic specimens to evaluate the variation in velocity.
Figure 17 Variation in velocity through the sides of the 13 weld cubes with respect to the
mean.
From early on it has been known that austenitic stainless steel welds develop a columnar
grain structure during solidification [3, 4]. These dendritic grains grow along the directions of
maximum heat loss during cooling. The elongated and oriented grains can grow typically up
to several millimetres in length. X-ray diffraction technique has been applied to obtain the
texture of the austenitic welds, and it was found that the long axis of the columnar grains
corresponds to the 100 crystallographic direction [3]. To establish the true propagation
directions of the sound it is necessary to know the elastic stiffness encountered by the wave.
Ref: TP AA058J-FR
Version: 01.05
To understand how the ultrasonic wave is affected by the microstructure of the weld, models
can be used but they need a quantified description of the weld. Since it is the sound velocity
that is of interest, then the elastic stiffness variation in the weld needs to be quantified. The
large dendritic grains can be observed using optical microscopes but the optical microscope
is unable to provide quantified information. X-ray diffraction can also be used to quantify the
weld to a limited extent, as large sampling volumes are required leading to a very low
resolution description of the weld.
Recent advances in Scanning Electron Microscopes (SEM) has led to the development of
Electron Back Scatter Diffraction (EBSD) techniques which are able to map the orientation of
the crystallographic structure on the surface of a metallic specimen to very high resolutions.
The data provided by the EBSD technique results in a map which is able to reveal the
constituent grain morphology, boundaries and orientations [5]. With the high resolution map,
the behaviour of the sound field can be studied better and array ultrasonic concepts (of
steering and focusing sound energy) can be better tailored to the weld.
3.2 EBSD approach to quantifying the weld
From the DISSIMILAR specimen three cross sections of the weld was taken, two from the
ends and one from the middle, as shown in Figure 18.
The two samples at the ends of the plate (Samples 1 and 3) were extracted at a distance
greater than 25mm from the ends of the plate to avoid any end effects from the welding.
Sample 1 is near Flaw 2 and Sample 3 is near Flaw 8. Sample 2, the middle sample, was
taken between Flaws 4 and 5. EBSD maps were then created of all three samples.
The specimens were mechanically polished to a surface finish of 1µm and silica colloidal was
used as the final polishing step to eliminate the surface deformation due to the mechanical
polishing. The prepared specimens were then loaded into the SEM at an angle of 70° from
the incident electron beam and facing a phosphor detector. The electron beam was then
raster scanned across the sample in a grid pattern with data collected at a specified
increment (the scanning resolution). When the electron beam strikes the tilted crystalline
sample, the diffracted electrons (following the Bragg condition [5]) form a characteristic
Kikuchi pattern on the florescent phosphor screen.
The Kikuchi pattern carries information of the crystal structure and orientation of the sample
region from which it was generated. When the beam is scanned in a grid across the sample
surface, crystal orientation will be measured at each sampling point. The crystal orientation is
calculated from the Kikuchi band positions. The positions of the Kikuchi bands are found by
using the Hough transform to convert the Kikuchi lines to points in Hough space. Using the
calibrated system, the angles between the planes that produce the Kikuchi bands on the
phosphor screen can be calculated, then the results will be compared with a list of inter-
planar angles for the analysed crystal structure to allocate the correct Miller indices to each
detected bands. This last step will allow calculation of the orientation of the crystal lattice in
terms of Euler angles with respect to coordinates fixed in the sample [5]. With modern
computing resources, the whole process for calculating the orientation data using the Kikuchi
pattern takes less than a few milliseconds.
Ref: TP AA058J-FR
Version: 01.05
Figure 18 Position of the three samples, referred to as Sample 1, 2 and 3.
The raw data obtained is expressed in terms of the Euler angles notation (φ1, φ and φ2) and
the rotation sequence of the unit (FCC) crystal with respect to the sample reference is shown
in Figure 19(a). However, the three angles (termed α, β and γ) to orient the crystal in the
weld when input to the model (described in Section 6) describe a different sequence of
rotation around the x, y and z axes, as shown in Figure 19(b). An algorithm written in
MATLAB was used to convert the Euler angles to the required angles with respect to the
specimen axes; the transformation matrix is given in Figure 20.
Ref: TP AA058J-FR
Version: 01.05
Figure 19 Description of the rotation sequence to orient the crystals using (a) Euler angles
(φ1, φ and φ2) and (b) around the specimen axes x, y and z (α, β and γ).
Figure 20 The transformation matrix g, is used to derive the specimen related angles (b)
given the Euler angles (a) output by the EBSD technique.
3.3 Optimisation of EBSD scanning parameters
A primary parameter of concern is the step size or scanning resolution, which impacts on the
efficiency and cost of implementing the EBSD approach to weld quantification. The aim is to
select the scanning resolution such that sufficient crystallographic information is collected in
the shortest possible time. Optical and SEM analysis show that the grains in the weld range
from a few micrometres to millimetres. Additionally, it is known that the propagation of the
ultrasound is sensitive to the ‘mean’ grain size distribution of the material. In particular,
attenuation is known to have a strong relationship to grain size and the frequency (ie
wavelength) of the ultrasound [6]. Hence it is important to establish the ‘significant’ grain size
and the grain boundaries which affect the propagation of the sound and capture this for input
to the model.
Ref: TP AA058J-FR
Version: 01.05
3.3.1 The minimum scatterer criterion
The attenuation in the material can be classified to fit three possible models and which of
these three models that fit a particular inspection case depends on the ratio of wavelength to
mean grain size [6]. The first of these models is termed the Rayleigh regime in which the
wavelength of the sound is much larger than the mean grain size. The second is termed the
stochastic regime where the wavelength is similar to the mean grain size and the third is
termed the geometric regime where the wavelength is much smaller than the grain size. If
the inspection system is in the stochastic or geometric regime, then the attenuation is
excessive and useful inspection will not be possible. Hence, ideally, it is required to choose a
sound frequency that places the system in the Rayleigh regime, ie the wavelength being
larger than the mean grain size. Equation [3.1] describes the Rayleigh model of attenuation.
(D, f ) aR D 3 f 4 [3.1]
Where α is the attenuation coefficient (dB/mm), D is the grain size (mm), f is frequency (MHz)
and aR is a function of the material anisotropy (dimensionless coefficient).
As the sound wave propagates in the metal, grains become scatterers of the energy (the
strongest component of attenuation in austenitic materials) when they become larger than a
one tenth of the wavelength λ; that is when D 10
[2]. Experimental methods can be used,
under certain assumptions, to show the size of the grains that fit the model given in equation
[3.1].
Figure 21 shows a cuboid extracted from within an austenitic weld, with typical compositions
to those used in the nuclear primary coolant circuit. All the dimensions of the cuboid are very
accurately machined and the two large sides are polished to a very smooth finish. Through
transmission techniques were then used to measure the attenuation through the weld metal
coupon; Figure 22 shows the experimental setup within an immersion tank.
Figure 21 Extraction of an accurately machined and polished coupon from the weld.
Ref: TP AA058J-FR
Version: 01.05
Figure 22 Through transmission experimental setup to measure the attenuation.
The attenuation is measured in the cuboid block at several frequencies. The thickness of the
block (in the direction of attenuation measurement) is then decreased by a small step (eg
5mm) and the attenuation ate several frequencies is measured again, such that such
measurements can be taken at a minimum of 5 step sizes. Figure 23 shows the through
transmission data, determining the difference in gain required to place the transmitted signal
at the same amplitude as the incident signal.
Figure 23 Through transmission data to evaluate the attenuation through the coupon.
A minimum of three frequencies (ie the centre frequency of the transmitting probe) should be
selected around the frequency at which the inspection is to be designed. The data presented
here was taken at frequencies of 1, 2.25 and 3.5MHz using well characterised immersion
probes. Figure 24 shows the attenuation vs. thickness graphs for the dataset at the three
frequencies. Assuming a linear relationship, a line of best fit is generated whose gradient is
equal to the attenuation coefficient α at that frequency.
Ref: TP AA058J-FR
Version: 01.05
Figure 24 Attenuation vs. the thickness at the three frequencies, showing lines of best fit.
Subsequently the graph of the attenuation coefficient vs. the frequency can be generated
(Figure 25) which shows the expected exponential increase in attenuation with increasing
frequency. However, note that the curve of best fit is a polynomial of order 2 and the use of
only three frequencies limits exploration of any higher order relationships.
Figure 25 Attenuation coefficient (α) vs. the thickness showing the curves of best fit.
A plot of the attenuation coefficient vs. frequency to the power of 4 (ie to f 4) then allows
assessment of whether the measured attenuation fits the assumed Rayleigh model. Figure
Ref: TP AA058J-FR
Version: 01.05
26 shows that the minimum three data points (at the three values of f) fit a linear relationship
assumed by equation [3.1].
Figure 26 Plot showing the fit of the data to the assumed Rayleigh model.
The gradient of the line of best fit in Figure 26 is then equal to the product of the parameter
aR and the grain size D3, according to equation [3.1]. The parameter aR is proportional to
another parameter Δ2, which is a function of the stiffness constants. The austenitic material is
known to be anisotropic and the stiffness constants of a material with a similar composition
were measured using a single crystal [7]. In a single crystal, all the lattice structure is
organised to align with the three major crystal axes of the unit crystal. Using ultrasonic
velocity the three basic elastic stiffness constants can be measured. Hence the constants are
C11=203.6GPa, C12=133.5GPa and C44=129.8GPa [7]. A discussion regarding the elastic
constants is presented in Section 3.7. Given that,
2C 44
1
C11 C12
, and the required constant is,
2
aR 7.3077 . Now assuming the constant of proportionality to be equal to 1, we get
2
the relation: D3 1.2325 .
This then gives a value of 0.55mm for the grain size. If the frequency of inspection is chosen
to be 1.5MHz, then the wavelength of the longitudinal wave would be ~3.8mm, which is
about 7 times greater than D; alternatively, D ~ 1/7th of the wavelength.
In summary, the experimental data supports the view that grains become significant
scatterers of the sound energy when they approach a size approximately 1/10th of the
wavelength [2]. This implies that the description of the weld inside the model must sufficiently
capture significant scatterers, but can neglect detail that is smaller (which would lead to
needless inefficiencies and increased cost). This is termed the minimum scatterer criterion
and is then used to set the scanning resolution for the EBSD scans.
Ref: TP AA058J-FR
Version: 01.05
3.3.2 Empirical study of scanning resolutions
Several trials were undertaken to optimise the scanning resolution (step size), electron beam
current, the video gain and the integration time of the EBSD camera. The aim of the studies
was to optimise parameters that maximise acquisition rates and reduce the overall cost of
scanning the entire weld sections. As stated earlier, a primary goal is to select sufficient
resolution for the model to effectively predict the sound propagation. Figure 27 shows the
orientation maps of an area 2.5mm x 2.5mm when using different scanning step sizes. The
best resolution showing well defined boundaries was given by the smallest step size of 5µm
as shown in Figure 27(d). However the acquisition time at a 5µm resolution is 60 times
longer than the scan with a step size of 40µm; considering that the area shown here is only
6.25mm2, the required scanning time for a full weld cross section will be excessively costly.
However, for the required purposes, the resolution of 40µm is deemed sufficient as it
captures the boundaries between the regions sufficiently well. Note that at a frequency of
1.5MHz the longitudinal wavelength is ~3.8mm with the minimum scatterer being 0.38mm.
Hence a 40µm step size fits ~10 sampling points within the width of the scatterer, which is
sufficient to capture the significant regions within the weld map. Hence based on the
minimum scatterer criterion a step size of 40µm was selected.
Figure 27 Scans at several step sizes and the corresponding time to completion.
3.4 Processing the EBSD data
Processing is necessary to convert a data set containing 3.8million sampling points covering
an 80mm by 86mm weld cross section into a simpler orientation map for input to the
modelling platform (Section 7). Determining the grain boundaries is dependent on the level of
misorientation allowed within a region deemed to be a grain, ie how much orientation
difference between areas was required to assign them as different grains. Allowing the
misorientation range to increase effectively coarsens the grains. Figure 28(a) and 28(b) show
the complexity of the orientation map when the allowed misorientation within a region defined
to be a grain is small. The number of grains reduces significantly when the misorientation
range rises to 20°, as shown in Figure 28(c).
Ref: TP AA058J-FR
Version: 01.05
Figure 28 Grain orientation map when the misorientation range is (a) 5°, (b) 10° and (c) 20°.
The area represented is about 20mm by 20mm at the clad, weld, buttering interface.
The misorientation within a grain or across grain boundaries can be quantified using the
EBSD data, as shown in Figure 29. The raw map makes use of colour gradient to represent
the misorientation within a grain. The insets in Figure 29 are the misorientation profiles over
the lines indicated by the arrows. However, the concept of misorientation within grains
cannot be transferred to the model. For the model, well defined regions (ie within closed
boundaries) and with each closed region assigned one particular orientation is required. This
gave rise to the development of a processing step termed orientation unification.
The aim of the orientation unification method, like the scanning resolution, is to satisfy the
minimum scatterer criterion. Firstly, select a misorientation range such that the smallest
grains approach the size of the minimum scatterer; this misorientation range for the
DISSIMILAR weld specimen was evaluated to be 20°. Secondly, statistical analysis over the
entire weld map was performed to select a small number of key (major) orientations around
which the misorientation range was placed. Hence, the weld map is now composed of
regions which are one of these major orientations and each major orientation is assigned its
own distinct colour.
Figure 29 Misorientation profiles along two lines, with the insets showing the misorientation
along the lines which is represented on the 2D map as a colour gradient.
Ref: TP AA058J-FR
Version: 01.05
Table 3.1 shows the eleven major orientations identified in the buttering and weld regions in
terms of the Euler angles. Table 3.2 shows the eleven major orientations in terms of the
three angles with respect to the specimen axes, converted using the transformation given in
Figure 20. Hence the entire weld can now be described in a map which makes use of a
number of major orientations to describe closed regions which satisfy the minimum scatterer
criterion.
Table 3.1 Euler angles of the eleven major orientations in the DISSIMILAR weld specimen.
Colour 1( ) ( ) 2( )
(rotation about z) (rotation about x) (rotation about z)
Red 0 0 0
Lime Green 103.8 15.9 68
Yellow 190.1 28.9 66.9
Blue 265.1 43.9 9
Fuchsia 310.9 7.9 11.1
Cyan 68.1 45 85.6
Maroon 44 23.9 14.5
Purple 322 39.3 0.6
Gray 231.8 32.29 29.93
Green 330.6 40 16.8
White 250.5 42.7 56.3
Table 3.2 The eleven major orientations expressed with respect to the specimen axes.
Colour ( ) () ()
(rotation about x) (rotation about y) (rotation about (z)
Red 0 0 0
Lime Green 176.1 195.4 352.3
Yellow 331.5 4.9 255.8
Blue 355.3 43.7 272.2
Fuchsia 5.1 6.0 322.3
Cyan 200.5 221 326
Maroon 197.7 196.3 235.8
Purple 32.8 23.0 329.4
Gray 338.6 24.8 257.0
Green 36.4 18.5 353.5
White 342.9 40 300.6
3.5 Analysis of the microstructure and texture of the austenitic weld
The V-prep weld section generated from the EBSD scanning and post-processing is shown
in Figure 30. In the weld, the dendritic grains display epitaxial growth, starting from the earlier
weld deposits growing upwards through subsequent deposits. The orientation of the
dendrites follows the heat flow directions. Note that the grains in the buttering region are near
horizontal and there is a clear difference between the orientations of the weld and the
buttering, as indicated by the colours. The major orientations in the weld are coded red, lime
Ref: TP AA058J-FR
Version: 01.05
green and blue, with orientations (001)[100], (011)[100] and (013)[100], respectively; note
that orientation component expressed in the form (hkl)//y, [uvw]//z). The weld metal exhibits a
strong [100] fibre with the [100] direction oriented along the z-axis, as shown in Figure 30.
This is consistent with the observed solidification growth in cubic crystalline materials. The
grains have sizes in excess of 2 to 4mm and are tilted from the z-axis by up to 35°.
The grains on either side of the austenitic weld have fibrous orientation which is near [-3-
10]//z or [-210]//z, as shown in Figure 30(d) and (f). The grains in the buttering are
perpendicular to the z-axis, ie they have a 90° rotation about the y-axis when compared to
the orientation in the weld region. The major orientations in the buttering are yellow
(101)[010] and red (001)[100]. Figure 31 shows a schematic of the 3D crystal orientations of
the eleven major orientations to allow visualisation of how the crystals exist in the different
regions of the weld.
Figure 30 The optical and EBSD processed weld maps along with a texture analysis of
different regions of the weld.
Ref: TP AA058J-FR
Version: 01.05
Figure 31 A schematic showing the orientations of the cubic unit crystal in the different
Figure 32 shows the three weld sections extracted from the DISSIMILAR specimen (as
shown previously in Figure 18) and they show that (1) they are well described by the same
major orientations and (2) the volume fraction of those major orientations is similar.
Figure 32 The orientation maps showing the eleven assigned orientations (see Table 3.1) of
(a) Sample 1 near Flaw 2, (b) Sample 2 in the middle and (c) Sample 3 near Flaw 8.
The key texture direction parallel to [100] (in red) found in all three weld sections is shown in
Figure 33. Samples 1 and 3 (ie those near the ends of the DISSIMILAR plate) exhibit similar
distribution of the [100] fibre, with the largest concentration in the middle of the V-prep weld.
However, in Sample 2 the distribution of the [100] fibre is pushed towards the stainless steel
parent. Hence, although the volume fraction of [100] fibre is similar in near the ends of the
plate and the middle, the distribution appears to be varying.
The distribution is determined by the process conditions (ie the heat flow directions) but this
can have an effect on the ultrasonic propagation. This then requires consideration of how the
Ref: TP AA058J-FR
Version: 01.05
weld microstructure varies parallel to the welding direction, ie in a direction perpendicular to

the cross sections shown in Figures 32 and 33.
Figure 33 The distribution of the [001] fibre in three weld cross sections samples; (a) Sample
1 near Flaw 2, (b) Sample 2 in the middle and (c) Sample 3 near Flaw 8.
3.6 Lengthwise uniformity
The study of the weld microstructure variation parallel to the welding line, what is now termed
the lengthwise uniformity, was undertaken using both EBSD analysis and an ultrasonic
technique using Flaw 1.
3.6.1 Analysis of lengthwise uniformity using EBSD
Four specimens were extracted from the weld as indicated in Figure 34. Specimens L2 and
L4 were scanned by EBSD while specimens L1 and L3 were examined using optical
microscopy. Each specimen was 25mm by 15mm by 4mm thick and the large side (ie 15mm
x 25mm) contains the microstructure which develops parallel to the welding direction, ie
along the y-axis; EBSD was then used to scan the surface.
Ref: TP AA058J-FR
Version: 01.05
Figure 34 Positions and dimensions of specimens extracted from the weld for lengthwise
uniformity analysis along the y-axis.
Specimen L2 was cut near the centre of the weld and Figure 35 shows the orientation map
using the major orientations (Tables 3.1 and 3.2). Hence, as in the cross section maps
shown in Figure 32, the major orientations of Table 3.1 are sufficient to fully map the
scanned y-z plane. The dendritic grains take the orientations red (001)[100], lime green
(011)[100] and blue (013)[100]. The smaller grains take the orientations near [-3-10]//z or [-
210]//z. The width of the columnar grains takes an average value of 650µm (with a scatter of
200µm) and the lengths greater than 10mm, which is consistent with the observations in the
x-z plane (ie the weld cross section maps).
Ref: TP AA058J-FR
Version: 01.05
Figure 35 Processed orientation map of specimen L2 using the major orientations given in
Table 3.1 and an optical image of specimen L1 close to the same region, identifying the inter-
run boundary shown in cyan box.
Specimen L4 was close to the bottom of the weld near the buttering region, ie close to the
edge of the weld. As shown in Figure 36, the grains at the top of the specimen are dendritic
with major orientations of blue (013)[100] (see corresponding region in Figure 32(a)),
whereas those grains near the bottom are smaller with average dimensions of 40µm (width)
by 100µm (length). The lower region on specimen L4 has the same orientations as the
buttering region identified in Figure 32, which are mainly yellow (101)[010] and red
(001)[100]; the grains in this region are equiaxed.
Ref: TP AA058J-FR
Version: 01.05
Figure 36 Processed orientation map of specimen L4 using the major orientations given in
Table 3.1 and an optical image of specimen L3 close to the same region showing the inter-
run boundary shown in cyan box.
Using the orientation maps of the samples and specimens extracted in the x-z and y-z
planes, the average grain size of the different regions has been evaluated and given in Table
3.3. Despite the large scatter in the data, the shape of the grains can be approximately
described as rods. The grains away from the edge of the weld are generally large and long,
whilst those near the weld edges are small, short and tilted. A schematic of the grain shapes
in the different regions is shown in Figure 37.
Table 3.3 Average grains sizes of the dendrites in different regions of the weld.
x-z plane (data taken from the y-z plane

Sample 1 in the regions close to
specimens L2 and L4).
W (µm) L (µm) W(µm) L(µm)
Buttering Width of the 780 ± 300 700 ± 300 650 ± 300

butter
Large grains 900 ± 300 10000 to 20000 648 ± 200 10000 to 20000
Small grains 62 ± 30 150 ± 50 40 ±30 100 ± 50
Ref: TP AA058J-FR
Version: 01.05
Figure 37 Schematic showing the rod like grains and their orientations in the different
3.6.2 Analysis of lengthwise uniformity using ultrasonics
Flaw 1 (Figure 6) was introduced into the DISSIMILAR specimen for the purpose of
ultrasonically studying the uniformity of the weld microstructure along the welding direction.
Several experiments were undertaken during the ultrasonic baseline scans (Section 7) and
the results were then analysed. Figure 38 shows the plan view of the inspection showing the
probe which travels parallel to the weld centreline at a fixed standoff while the signal from the
corner and tip of surface breaking slot Flaw 1 was monitored.
Figure 39 shows the inspection in side view and the data sets from the automated baseline
inspection that was used for the uniformity analysis. The automated baseline was chosen as
it contains encoded data with well-defined stand-off distances with respect to the datum.
Additionally the frequency of the probe was 1MHz and the beam angles chosen were 45, 60
and 65°. The data from the austenitic side did not require the sound beam to traverse the
weld; hence it is the reference data set.
Ref: TP AA058J-FR
Version: 01.05
Figure 38 Plan view of the lengthwise uniformity experimental setup.
Figure 39 Side view of the lengthwise uniformity experimental setup, probes used and the
data sets from the automated baseline scans.
Figure 40 shows the data from the reference scans performed from the austenitic side. The
position of the artificially introduced flaws (Figures 5 - 13) is shown along the y-axis direction
Ref: TP AA058J-FR
Version: 01.05
in Figure 40 (and the following figures) to aid the analysis of the data. Figure 40 shows that
the variation in amplitude of the slot corner over the length of the weld does not change by
more than ~2dB for the two beam angles. Note that the vertical dashed line in Figure 40 at a
y-axis position of 290mm represents the point where the as-received DISSIMILAR specimen
(Figure 14) was halved to take the middle section for EBSD mapping.
Figure 40 Variation in the signal amplitude from the slot corner measured from the austenitic
side (reference scan).
Figure 41 shows the variation in the horizontal position of the slot corner measured from the
austenitic side. In this dataset the effects due to the ends of the plate are more pronounced
with variation in excess of 4mm near y=0mm and y=290mm. Figure 42 shows the variation in
the vertical position of the slot corner measured from the austenitic side.
With the probe positioned on the ferritic side, the sound beam has to travel through the weld
to approach the slot. Hence the sound wave will be subject to the distortive effects of the
weld. Figures 43 to 45 show the data collected with the sound travelling through the weld.
Figure 43 shows variation in the amplitude of the slot corner, Figure 44 shows the variation in
horizontal position of the corner echo and Figure 45 shows variation in the vertical position of
the corner echo.
Ref: TP AA058J-FR
Version: 01.05
Figure 41 Variation in the horizontal position of the slot corner measured from the austenitic
Figure 42 Variation in the vertical position of the slot corner measured from the austenitic
Ref: TP AA058J-FR
Version: 01.05
Figure 43 Variation in the signal amplitude of the slot corner measured from the ferritic side.
Figure 44 Variation in the horizontal position of the slot corner measured from the ferritic
side.
Ref: TP AA058J-FR
Version: 01.05
Figure 45 Variation in the vertical position of the slot corner measured from the ferritic side.
Note in Figure 44 that the data from the 1MHz 45° beam for y>290mm is significantly offset
in comparison to the other datasets. The reason for this discrepancy is not known, however
the variation to itself is similar to the other datasets.
For each flaw, the error between the measured through-wall positioning of top and bottom
edges of the flaws (when using the conventional automated technique described in Section
6.3.2 and reported in Tables 3 to 9 of Appendix E) and the nominal values (confirmed
through sectioning for Flaws 2, 3 and 4) is presented in Figure 46. This data indicates
changes in the measurement error along the length of the weld, which could potentially be
related to changes in the microstructural condition.
Figure 46 Error in the through-wall position of flaw edges using the conventional automated
technique (see Section 6.3.2) compared to nominal values.
Ref: TP AA058J-FR
Version: 01.05
3.6.3 Summary of lengthwise uniformity analysis
The analysis using EBSD shows that, over the 15mm studied, the microstructure is
fairly uniform in the lengthwise direction.
The ultrasonic analysis shows that the overall variation in positional changes is within
2mm, with the maximum variation being 5mm.
The greatest effect of the weld anisotropy appears to be on signal amplitude, which is
nevertheless within 2dB.
The uniformity study was aimed at determining whether sampling the weld microstructure (by
taking several cross sections) for mapping is valid. In conclusion, the measured uniformity of
the microstructure along the welding direction appeared to suggest that this weld was
relatively uniform and hence the sampling approach was likely to be valid for the purposes of
sufficiently quantifying the weld within the models.
3.7 Evaluation of the stiffness constants
To be able to model the sound propagation in the austenitic weld, the stiffness constants of
the cubic crystalline system must be evaluated as accurately as possible. A popular method
of doing so experimentally is the use of single crystals of that alloy and measuring the
ultrasonic velocity of longitudinal and the two polarized shear wave modes [7, 8]. The
velocities of the different sound wave modes are related to the three independent stiffness
coefficients (C11, C12 and C44) of the cubic system, which can then be derived.
In the DISSIMILAR project an attempt was made to grow a single crystal using consumables
of the Inconel 182 specified for the Manual Metal Arc (or SMAW) process specified for the
specimen. Casting methods were used in the attempt to grow the crystal, however
subsequent SEM analysis showed sufficient uniformity was not achieved.
Nickel super alloys are now widely used for aerospace applications (in particular for single
crystal turbine blades) and extensive information regarding these super alloys (eg CMSX486,
CM186LC, Rene N4) is available in the public domain [8, 9, and 10]. In the case of
aerospace materials, researchers have noted that there is little impact on the composition of
alloys on the measured stiffness coefficients [8, 9]. Based on this assumption, the values of
the stiffness coefficients measured earlier [7] on an austenitic alloy with similar composition
to the alloy used in the weld of the DISSIMILAR specimen was used to map the weld and
input to the model for studying the propagation of sound waves.
It is noted that since the exact values of the stiffness in the DISSIMILAR weld was not
determined, the subsequent work on adapted delay laws (Section 6) will have a degree of
error. This error will be manifest in the experimental validation data and will be considered in
the performance analysis of the techniques (Section 7).
The general recommendation for future effort is to, where the intention is to make use of
advanced techniques such as adapted delay laws (Section 6), undertake the growth of a
single crystal of the alloy to be used in the weld. Growth of single crystals is not easy and
several different methods to grow the crystals may need to be investigated before success.
However, establishing the actual stiffness coefficients will allow for better technique design.
Ref: TP AA058J-FR
Version: 01.05
3.8 Summary of EBSD analysis
The EBSD analysis showed that crystalline growth during the solidification of the weld
led to a macroscopic texture containing epitaxially grown dendrites along the
preferred fibre direction of (001)//z.
The scanning parameters, in particular the resolution, must be set according to the
minimum scatterer criterion in order to minimise costs.
Processing using the orientation unification concept has allowed preparation of the
EBSD data for input to the model, with closed regions (using boundaries) that are
assigned a major orientation containing the range of allowed misorientation.
Analysis of the lengthwise microstructure shows good uniformity along the welding
direction.
Results from the x-z and y-z planes suggest that the grains are rod like in shape,
consistent with previous experience which found dendritic growth.
It is possible that the sound wave may not be overly sensitive to some of the major
orientations, which were selected statistically. Validation efforts to investigate the influence of
particular orientations with respect to a propagating sound wave may allow for further
simplifications when mapping the weld. However, such refinement work was not undertaken
in the DISSIMILAR project and remains a task for future efforts.
3.9 Financial costs of implementing EBSD
EBSD scanning is implemented using a special detector attached to a Scanning Electron

Microscope (SEM). EDAX (www.edax.com) and Oxford Instruments (www.oxinst.com) are
leading suppliers of SEM and EBSD instrumentation. At the time of writing this document, the
estimated cost of a standard SEM and EBSD system is £300,000. In addition specialist
training will be required of the operator who will require a firm basis in metallurgy and
knowledge of operating the SEM system.
For an organisation interested in acquiring the capability to implement EBSD analysis in-
house, the primary cost will be the acquisition of the equipment and personnel costs.
Additionally there will be costs associated with regular calibrations, possible costs for repairs
to the systems, machining costs, chemicals for specimen preparation and overheads. An
alternative to maintaining the capabilities in-house would be to subcontract the EBSD
scanning and analysis to third parties. Such third parties are likely to be institutions (such as
TWI who possess SEM/EBSD facilities) or universities (such as the University of Birmingham
who undertook the microstructural development work in the DISSIMILAR project).
Regardless of which method is chosen, quantification of the weld microstructure using the
EBSD technique involves considerable expense. In the DISSIMILAR project three sections
were extracted and fully scanned at a cost to the project of £113,000. Future attempts may
require the scanning of more sections, different materials (involving further investigations into
specimen preparation), changes in parameters (resolution, misorientation) and further
exploration of implementation methods, which may lead to additional costs. Choosing to
explore the methods and techniques developed in the DISSIMILAR project is therefore not
an easy option. Therefore this approach is only likely to be viable for industries (such as the
power and oil & gas) where extremely critical components are present, where the
consequences of failure would adequately justify such costs to mitigate.
Ref: TP AA058J-FR
Version: 01.05
4 Probes
Ultrasonic transducers or probes make use of the piezoelectric effect to both generate and
detect ultrasonic energy [2]. The aim of the DISSIMILAR project was to make use of
advanced materials and methods available at the time to manufacture and then investigate
the implementation of array probes for the inspection of the DISSIMILAR DMW. The
transducer technology was provided by Alba Ultrasound of Glasgow, UK.
Two types of arrays were specified and manufactured for the DISSIMILAR project and their
performance was compared with other probe types used in three different baseline
inspections (Section 7). In this section, the methods and models used to specify the two
DISSIMILAR array probes is presented. The approach to specification illustrated in Sections
4.2, 4.3 and 4.4 provides information on the typical steps involved in undertaking
specification of an ultrasonic array for the inspection of austenitic welds.
4.2 Requirements
The difficulties of inspecting austenitic welds have been outlined earlier. In general, for thick
section (ie >25mm) welds several key requirements are known:
1. Longitudinal wave mode must be used

2. Input power (ie pulsing voltage) applied to probe must be sufficient
3. The ultrasonic pulse generated must contain a broad frequency content
4. Low frequency is required to reduce the levels of attenuation
Additional requirements for effective implementation of array techniques involving focusing of

the sound energy include ensuring that the radiating area of the probe is large enough for the
near zone length to be greater than the range at which focusing is required.
The use of longitudinal waves is recommended from experience when the full volume of the
weld needs to be inspected. Supplemental techniques involving the use of shear waves
skipping (reflecting) on the back wall of the specimens maybe used for detection of flaw such
as lack of side wall fusion (ie not requiring the sound to enter the weld), but such techniques
will not be discussed in this document. Similarly, experience shows that the instrumentation
must be able to provide sufficient power for the sound wave to be able to penetrate sufficient
distance within the weld, ie the pulsing voltage must be greater than 200V (Section 5).
The bandwidth, ie the frequency content in the sound wave, is required to be as broad as
possible, such that the sensitivity to a wide range of discontinuities is increased. In contrast,
the requirement for low frequency is a balance between required inspection range and
attenuation. As discussed in Section 3, the wavelength is required to be much larger than the
mean grain size along the propagation path to place the system in the Rayleigh regime of
attenuation. If the welded component was in existence then the first task would be to trial
several frequencies to establish sufficient penetration vs. attenuation. However, in the case
of the DISSIMILAR project the specimen was not in existence when the probe specification
effort started and hence a decision on frequency (ie based on experience) had to be taken.
The chosen frequency was 1.5MHz which is known to be well optimised for inspection of the
welds fabricated using the procedures used for the DISSIMILAR specimen previously.
Ref: TP AA058J-FR
Version: 01.05
4.3 Specification approach
Specification of ultrasonic probes through modelling is inherently an iterative process that

may take several iterations to arrive at the final values of key parameters. In this document,
the process undertaken and the results of specifying the two arrays are presented.
The first level specification pertains to ensuring that the probe will have the required
ultrasonic characteristics, ie sufficiently configured beam size at the required position to allow
for sizing, along with sufficient focusing to enhance probability of detection with good signal-
to-noise (S/N) characteristics. The second level specification regards the manufacture of the
array, ensuring that the materials are chosen and configured sufficiently well to output the
specified performance of the first level specification. The first level specification was
undertaken using a model termed SimulUS developed by Peak NDT of Derby [11, 12] and a
model termed CIVA [13] developed by the Commissariat à l’Énergie Atomique (CEA) of
France. The second level manufacturing specification was undertaken using a model termed
PZ Flex developed by Weidlinger Associates Inc. of USA. The engines of the three models
explored in the DISSIMILAR project are very different:
1. SimulUS uses Huygens’ Principle to plot the sound field ahead of a radiator.
2. CIVA uses semi-analytical methods to plot the sound field and evaluate interaction of
the field with discontinuities (ie flaws) in the medium.
3. PZ Flex is a finite element model with a well refined engine based on fundamental
wave propagation equations and is focused on modelling the piezoelectric effect.
All three models have been validated for the particular task they were used for in this
specification effort and examples of using all three models will be presented in this section.
4.4 Specification of the TRL-1 array
The final specification of the TRL-1 array is given in Appendix B. SimulUS and CIVA were
used for the first level specification and PZ Flex was used for the second level specification.
4.4.1 First level specification (CIVA and SimulUS)
4.4.1.1 Approach
Aim is to select values for parameters in order to provide the best possible performance for
the detection of flaws and sizing them in wall thicknesses up to 85mm. In specifying the
probes the assumption is that the material is isotropic, ie the effects of the austenitic weld
structure are not taken into account; the impact of the austenitic weld will be accounted for in
the design of the technique, making use of the ability of arrays to manipulate the sound field
ahead of them. However, it is important to ensure that the probe characteristics are
optimised for effective control of the sound beam within the volume of the weld.
The TRL-1 probe has a specific configuration known as Transmit-Receive Longitudinal

(TRL), where the transmitting and receiving arrays are separated such that the setup is
essentially in a pitch-catch configuration. The array is further configured to generate the
longitudinal wave mode. In order to avoid mode conversions when reflecting on boundaries
(such as the back wall), all inspection techniques will be assumed to be in half skip mode (ie
direct incidence). A TRL probe is made of transmission (Tx) and reception (Rx) parts as
shown in Figure 47.
Ref: TP AA058J-FR
Version: 01.05
(a) (b)
Figure 47 TRL profile along the (a) primary and (b) secondary axes.
In Figure 47 parameter Lp is the aperture of the array along the primary axis, equal to the
number of elements along the primary axis (np) x the element pitch (pp) and parameter Ls is
the aperture of the array along the secondary axis, equal to the number of elements along
the secondary axis (ns) x the element pitch (ps). Separating transmit and receive paths leads
to the better S/N performance when applied to austenitic welds [14]. Key advantages given
by the TRL configuration are:
Elimination of the probe to steel interface echo.

Providing a diverse beam path through the material providing some grain structure
noise averaging.
Eliminates the beam dead zone enabling the near surface material to be examined.
Table 4.1 is a summary of the parameters which are part of the optimisation effort.
Table 4.1 Probe parameters which are chosen and those that need to be optimised.
Parameter Part of Initial value @ Iteration 1

optimisation?
Probe type NO TRL
Wave mode NO Longitudinal
Array type YES 2D
Inspection type NO Direct / half skip
No. of elements on primary axis (np) YES 16
Primary axis pitch (pp) YES 5mm
No. of elements on the secondary axis (ns) YES 4
Secondary axis pitch (ps) YES 5mm
Frequency NO 1.5MHz
Coupling material NO Water (immersion)
Inclination / wedge angle (i) YES 11.2°
Roof angle (r) YES 5°
Inter array space (l) YES 6mm
Coupling path distance YES 15.6mm
The primary route to specifying the probe is to consider the sound field or sound beam
generated by the probe, in particular the shape and energy in the beam. Modelling allows the
evaluation of main beam characteristics at different steering angles and focus depths, and
the amplitude of unwanted echoes around this main beam. To quantitatively compare the
Ref: TP AA058J-FR
Version: 01.05
simulated beams the -6dB cross beam size at the focus range is used. In order to achieve
the best sizing resolution, the -6dB cross beam size must be minimised which also increases
the intensity of the sound (ie focusing). Additionally, the optimisation will aim to eliminate side
energy lobes which can lead to increases in the energy. The optimisation was required to
provide good beam characteristics from a depth of 5 to 85mm. The optimisation effort
underwent four iterations in total, which are presented and analysed in this document.
4.4.1.2 Iteration 1
The starting parameters values are given in Table 4.1. Figure 48 shows the beam cross
section and side views along the primary and secondary axes as computed using CIVA.
Computation
area
Computation
area
(a) Longitudinal wave cross beam section. (b) Shear and longitudinal modes in the
primary axis plane.
Computation
area
(c) Field in the plane of the secondary axis
Figure 48 Configuration of views generated in CIVA.
Four cases were investigated as part of the optimisation study:
1. Beam steered at 40 , focused at 5mm deep.

The beam focused at 5mm was to ensure coverage of near surface and that at 85mm was
aimed at finding flaws near the back wall. The steering range required was from 40 to 70°;
the sound field is required to maintain integrity at both extremes of this range.
Figures 49 to 52 show the results computed by the model for the longitudinal wave mode.
Ref: TP AA058J-FR
Version: 01.05
Beam steered at 40 Beam steered at 70

5mm 5mm
5mm deep
Focus at
85mm 85mm
85mm deep
Focus at
Figure 49 Plot of beams computed in the primary axis plane for 40 and 70° beams at focal
depths of 5 and 85mm.

L and T wave L wave L and T wave L wave
Focus at 5mm
deep
10mm 10mm 10mm 10mm
L and T wave L wave L and T wave L wave

Focus at 85mm
deep
10mm 10mm 10mm 10mm
Figure 50 Results for cross beam section computation for 40 and 70° beams at focal depths
of 5 and 85mm.
Ref: TP AA058J-FR
Version: 01.05
Figure 51 The -6dB beam size along the primary axis and the secondary axis for different
beam steering angles and focal depths.
Figure 52 The beam strength for 40 and 70° beam angles at focal depth of 5 and 85mm.
Ref: TP AA058J-FR
Version: 01.05
At 70°, the beam is not able to focus at a range of 248.5mm (equivalent to 85mm deep)
because the natural near zone range is 122mm and it is not possible to focus at a range
greater than the natural near zone of the probe. In addition, the 70° beam width is greater
than 50mm wide, with high sub-surface interference and a decrease of the beam strength by
25dB in comparison with a beam steering at 40 focused at 85mm.
Figure 53 shows the results of a computation to visualise the effect of reducing the width of
the probe to a single row of elements 5mm wide (ie a linear array). Figure 53 shows the
results for a linear array with 16 elements along the primary axis with 5mm wide elements. It
was not possible to achieve good focusing capability to a depth of 85mm at 70° and
consequently the size of the probe along the primary axis needed to be increased.
Cross beam section Side view along primary axis plane

85mm
10mm
Figure 53 Plot of the computed beam cross section and beam side view along the primary
axis for the equivalent linear probe (16 elements along the primary axis, element pitch 5mm
and element width 5mm) steered at 70˚ and focused at 85mm deep.
In summary, after iteration 1, the conclusions were:
1. A small beam size was achieved at short range with a beam size less than 5mm
along the primary beam axis and less than 10mm along the secondary axis.
2. A strong transverse wave occurs when the ultrasonic beam is focused near the
scanning surface. Propagation of transverse waves into the material can lead to an
increase in backscattered noise from the austenitic grain structure.
3. An aperture of 16 elements at 5mm pitch and 1.5MHz frequency cannot focus beyond
the maximum range of 120mm at 70°.
4.4.1.3 Iteration 2
A new element configuration needed to be used in order to reduce the creation of shear
wave modes, reduce interference near surface and, most importantly, improve the capacity
to focus at the required range at high beam angle. In this iteration, the effect on the ultrasonic
beam due to the change in the number of elements and the element pitch along the primary
and the secondary axes was investigated.
For every case the beam is steered at 70 and focused at a depth of 85mm. Iteration 2 was
undertaken using the SimulUS model. The key parameters kept fixed in Iteration 2 are given
in Table 4.2. In the DISSIMILAR project the maximum number of channels available in the
array controller (Section 5) was limited to 128. Hence transmit and receive halves had 64
elements each, and the aim of Iteration 2 was to find the optimum 2D configuration on each
half and the cases considered are listed in Table 4.3.
Ref: TP AA058J-FR
Version: 01.05
Table 4.2 Parameters kept fixed during Iteration 2.
Parameter Value
Probe type TRL
Wave mode Longitudinal
Array type 2D
Frequency 1.5MHz
Coupling material Water
Inclination / water wedge angle 13.1
Roof angle 2.5
Inter array space 11mm
Coupling path 27mm
Table 4.3 List of cases considered in Iteration 2.
Case Number of Number of Element Element Natural Natural

elements elements pitch pitch near zone Near field
in the in the in the in the length at length at
primary secondary primary secondary 0° 70°
axis axis axis axis (mm) (mm)
1 16 4 4 4 251 78
2 16 4 5 5 392 122
3 32 2 2 2 254 75
4 32 2 3 3 571 169
5 32 2 4 4 1016 300
6 32 2 5 5 1587 469
7 64 1 2 5 1587 303
8 64 1 2 10 1012 299
9 64 1 3 3 2315 689
10 64 1 3 5 2307 685
11 64 1 3 10 2291 678
12 64 1 3 20 2267 673
An array is not able to focus beyond the natural near field zone. The natural near zone length
is a function of the aperture (A) and the wavelength ( ) as given in Equation 4.1.
A2
Nz
4 [4.1]
The maximum required range is equal to 248.5mm in the material when the beam is steered
at 70 and focused at the backwall of the specimen (85mm). Hence aperture of the probe
has been chosen to obtain a natural near zone length greater than 248.5mm (see relevant
cases given in Table 4.3).
The results of all the cases in Table 4.3 is presented in Figure 54, showing the three views
necessary to visualise the beam generated by the probe as illustrated in Figure 48: beams in
the primary and secondary axes and the cross beam size in a plane at the range of the focal
position (85mm deep). Figure 55 shows the predicted beam sizes for all the cases.
Ref: TP AA058J-FR
Version: 01.05
Beam along primary axis Beam along secondary Beam size at

plane axis plane focus point
Secondary
Secondary axis
Case 1 Focu axis
16x4 s
4x4mm plan
Primary Primary Primary
axis axis axis
Case 2
16x4
5x5mm
Case 3
32x2
2x2mm
Case 4
32x2
3x3mm
Case 5
32x2
4x4mm
Case 6
32x2
5x5mm
Case 7
64x1
2x5mm
Ref: TP AA058J-FR
Version: 01.05
Case 8
64x1
2x10mm
Case 9
64x1
3x3mm
Case 10
64x1
3x5mm
Case 11
64x1
3x10mm
Case 12
64x1
3x20mm
Figure 54 Modelling results from SimulUS (-20dB threshold applied) for different array
configurations with a beam focused at 85mm deep and steered to 70 .
Ref: TP AA058J-FR
Version: 01.05
Figure 55 The -6dB cross beam measurement from the SimulUS models for all 12 cases.
In summary, after Iteration 2, the conclusions were:
1. In all cases, an increase in the aperture of the probe in primary or secondary axes
leads to a decrease in the beam size and an increase in the intensity of the sound at
the focus point. However, there can also be an increase in the side energy lobes
around the main beam.
2. For the same aperture size, an increase in the number of elements allows better
control of the beam and the energy along the beam axis has better distribution.
Moreover, by increasing the number of elements, the generation of side energy lobes
is reduced.
3. The use of several elements along the secondary axis allows the beam to be
controlled better. Four elements can steer and focus the beam. Two elements allow
the beam to be steered only. No control is possible with a single element. With the
linear array 64x1 elements, the beam will be controlled only along the primary axis.
The recommendations from Iteration 2 are:
1. Use minimum of 32 elements along the primary axis.

2. Using 2 elements along the secondary axis limits the steering and focal capability but
was deemed to be a satisfactory compromise in the particular case of the
DISSIMILAR project. The ideal solution would be to increase the total number of
channels available in the array controller.
3. Hence the array to be selected and taken into the next Iteration is the 32x2
configuration with a pitch of 5mm x 5mm.
Ref: TP AA058J-FR
Version: 01.05
4.4.1.4 Iteration 3
Iteration 3 is to assess the performances of the selected array (32x2 configuration,

5mmx5mm pitch) at the beam angles 40˚ and 70 , and at focusing ranges equivalent to
depths of 10mm, 20mm and 85mm deep. Table 4.4 shows the parameters kept fixed during
Iteration 3.
Table 4.4 Probe and inspection parameters kept fixed during Iteration 3.
Parameters Value
Probe type TRL
Wave mode Longitudinal
Array type 2D
No. of elements on primary axis (np) 32
Primary axis pitch (pp) 5mm
No. of elements on secondary axis (ns) 2
Secondary axis pitch (pp) 5mm
Frequency 1.5MHz
Coupling material Water
Inclination 13.1
Roof angle 2.5
Inter array space 11mm
Coupling path 27mm
In this iteration, an apodisation curve, shown in Figure 56 was applied to investigate the
possibility of reducing the effects of side energy lobes. Apodisation is where the excitation
energy (voltage) of each element is not uniform along the primary axis (rows) of the array
with the maximum energy in the centre of the array. Figures 57 to 63 present the results
generated using CIVA in Iteration 3.
Figure 56 The apodisation curve applied to each of the 32 elements on transmit and receive
rows.
Ref: TP AA058J-FR
Version: 01.05

10mm
10mm
10mm deep
Focus at
20mm 20mm
20mm deep
Focus at
85mm 85mm
Focus at 85mm deep
Figure 57 Beam computation results in the primary axis plane without apodisation.
Ref: TP AA058J-FR
Version: 01.05

L wave L wave
Focus at 10mm
deep
10mm 10mm 10mm 10mm
L wave L wave
Focus at 20mm
deep
10mm 10mm 10mm 10mm
L wave L wave
Focus at 85mm
deep
10mm 10mm 10mm 10mm
Figure 58 Results of the beam size at the focal positions without apodisation.
Ref: TP AA058J-FR
Version: 01.05

10mm
10mm
10mm deep
Focus at
20mm 20mm
Focus at
20mm
deep
85mm 85mm
Focus at 85mm deep
Figure 59 Beam computation results in the primary axis plane with apodisation.
Ref: TP AA058J-FR
Version: 01.05

L wave L wave
Focus at 10mm
deep
10mm 10mm 10mm 10mm
L wave L wave
Focus at 20mm
deep
10mm 10mm 10mm 10mm
L wave L wave
Focus at 85mm
deep
10mm 10mm 10mm 10mm
Figure 60 Results of the beam size at the focal positions with apodisation.
Figure 61 The -6dB beam size measurement from modelling along the primary axis with and
without apodisation curve.
Ref: TP AA058J-FR
Version: 01.05
Figure 62 The -6dB beam size measurement from modelling along the secondary axis with
and without apodisation curve.
Figure 63 Beam strengths for different beam steering angles and focus depths with and
without apodisation curve.
Ref: TP AA058J-FR
Version: 01.05
1. This probe configuration has eliminated the strong shear waves at short focal ranges.
2. The small focused beam size of 5mm along the primary axis and smaller than 10mm
along the secondary axis is modelled to be theoretically possible.
3. The model has shown that by using this element configuration the high energy near
the scanning surface has been reduced when the beam is focused at long range
(equivalent to 85mm deep).
4. Some side energy lobes are still present around the main beam at short and long
focusing ranges, and at both low and high steering angles.
5. At a focal depth of 85mm the beam size is now smaller than 5mm at 40 . At 70 the
beam has been significantly reduced from 34mm to 14mm.
6. The reduction in beam size along the primary axis at long focusing range has been
earned at the expense of beam size along the secondary axis: the beam size along
the secondary axis has increased by 14mm.
7. Modelling has shown that apodisation provides advantages and disadvantages. In
general, the beam shape is better with fewer side lobes but generates a stronger
shear wave. Additionally, at the focus point, the beam size is increased and the beam
strength is reduced.
8. Hence based on the evidence generated through modelling in Iteration 3, the use of
apodisation is not recommended. The reasons are firstly it does not lead to any
significant changes in beam sizes but it does lead to a significant reduction in beam
strength.
4.4.1.5 Iteration 4
The array 32x2 elements, 4x5mm pitch and 32x2 elements, 4x6mm pitch at the beam angles
40˚ and 70 and focusing ranges 10mm and 80mm deep were investigated. Table 4.5 shows
the parameters fixed in Iteration 4 and Figures 64 to 71 present the results.
Table 4.5 Probe and inspection parameters kept fixed during Iteration 4.
Parameters Iteration 3a Iteration 3b

Wave mode Longitudinal Longitudinal
Array type 2D 2D
No. of elements on primary axis (np) 32 32
Primary axis pitch (pp) 4mm 4mm
No. of elements on secondary axis (ns) 2 2
Secondary axis pitch (pp) 5mm 6mm
Frequency 1.5MHz 1.5MHz
Coupling material Water Water
Inclination 13.1 13.1
Roof angle 2.5 2.5
Inter array space 11mm 11mm
Coupling path 27mm 27mm
Ref: TP AA058J-FR
Version: 01.05

10mm 10mm
Focus at
10mm
deep
85mm 85mm
Focus at
85mm
deep
Figure 64 Beam computation in the primary axis plane for the array 32X2 4mmX5mm array.

10mm 10mm
Focus at
10mm
deep
85mm 85mm
Focus at
85mm
deep
Figure 65 Beam computation in the primary axis plane for the array 32X2 4mmX6mm array.
Ref: TP AA058J-FR
Version: 01.05

L+T wave L+T wave
Focus at 10mm deep
25mm 25mm
L+T wave L+T wave

Focus at 85mm deep
25mm 25mm
Figure 66 Results of the beam size at the focal positions for the 32x2 4mmx5mm array.

L+T wave L+T wave
Focus at 10mm deep
25mm 25mm
L+T wave L+T wave

Focus at 85mm deep
25mm 25mm
Figure 67 Results of the beam size at the focal positions for the 32x2 4mmx6mm array.
Ref: TP AA058J-FR
Version: 01.05
Figure 68 The -6dB beam size measurement from modelling along the primary and
secondary axes for the 32x2 4mmx5mm array.
Figure 69 Beam strength for different beam steering angles and focus depths for the 32x2
4mmx5mm array.
Ref: TP AA058J-FR
Version: 01.05
Figure 70 The -6dB beam size measurement from modelling along the primary and
secondary axes for the 32x2 4mmx6mm array.
Figure 71 Beam strengths for different beam steering angles and focus depths for the array
32x2 4mmx6mm.
Ref: TP AA058J-FR
Version: 01.05
1. Both configurations (32x2, 4mmx5mm and 32x2, 4mmx6mm) appear to make

possible a small focal beam size of 6mm along the primary axis and smaller than
10mm along the secondary axis.
2. Both configurations show a reduction of side energy lobes around the main beam at
long focused range.
3. With 32x2 elements, 4mmx6mm pitch, the shear wave appears to be reduced at long
range and at high beam.
4. With 32x2 elements, 4mmx6mm pitch, at a focal depth of 85mm the beam size along
the primary axis is increased by 4mm at 70 . However the beam size is still smaller
than 20mm in this direction. Along the secondary axis, by increasing the aperture the
beam width is now reduced by 5mm.
5. The beam strength produced by the 32x2 configuration, 4mmx6mm pitch is greater
by 4dB at 40 and at a focus point 85mm deep. At higher beam angle, this
configuration appears to produce the same beam strength level as the previous
configuration in Iteration 3 (32x2 5mmx5mm).
4.4.1.6 Discussion
Position and definition of computation area
The cross beam sections at the focal positions computed using CIVA and SimulUS were at
the intersection point of transmit and the receive beams. In all models, only the contribution
of the transmit beam is computed (free field). The beam computed at the intersection point
from the transmitter is the one that will interact with flaws to generate echoes for the receiver.
To completely model the pulse-echo system, the simulation of a target must be performed at
the desired range using the modelled probe.
Manufacturing tolerance
The tolerance in the manufacture of the probe is critical especially with regard to the wedge
and roof angles. These angles must be identical for both arrays in order to ensure the
intersection of transmit and the receive beams at the focus point defined by the delay law. In
particular an error of 0.1˚ in the wedge angle will generate a positional error of 1.8mm at
254mm range. Therefore the tolerance on each on tilt and roof angle needs to be better than
0.05˚ or a maximum difference between the two array halves of no more than 0.1˚. The
housing design for TRL-1 is presented in Section 4.3.3.
Apodisation
Apodisation is a capability available in the instrument (Section 5). The specification of the
probe can be done without apodisation but it can be applied at a later date if needed.
After Iteration 4, the review indicated that the specification of the array arrived at was
sufficient to perform the task of inspecting the DISSIMILAR DMW. Hence the specification
document in Appendix B was produced and submitted for manufacture, which then initiated
the second level specification, which considered the issued specification and aimed to select
the materials and manufacturing processes to achieve the required performance.
Ref: TP AA058J-FR
Version: 01.05
4.4.2 Second level specification (PZ Flex)
The aim of the second level specification is to ensure that the ultrasonic performance
demanded from the first level specification can be attained by the physically manufactured
probe. Of primary concern is the choice of the active material, ie the generation and
reception of ultrasonic energy. Secondary considerations include the generation of unwanted
modes which travel laterally in between the elements of the array. The piezocomposite
material, shown in Figure 72, is an advanced composite material which was used to
manufacture the array. The advantages provided by the piezocomposite materials over
traditional monolithic piezoceramics are:
1. Good suppression of parasitic resonances - ensure only the compression mode is

transmitted into the medium.
2. Increased sensitivity and efficiency - increased S/N and penetration
3. Increased bandwidth - for higher resolution imaging
4. Reduced cross talk
Figure 72 Structure of the 1-3 piezocomposite material, showing the active ceramic pillars
embedded in a soft polymer matrix.
A number of lateral wave modes could be setup in the piezocomposite which could then
interfere with the thickness mode (which is the one of interest); these different modes are
illustrated in Figure 73. The aim of the specification is to ensure that either these modes are
well separated from the frequency of interest or are damped out by careful choice of
materials.
The second level specification procedure used was to (1) use the requirements specified in
the first level specification (eg sensitivity, bandwidth, cross talk etc) and (2) use the finite
element model PZ Flex to determine the microstructure (ie design out the lateral resonances
and meeting the requirements of bandwidth, sensitivity etc).
Ref: TP AA058J-FR
Version: 01.05
Figure 73 Unwanted lateral modes which could exist in the piezocomposite material.
PZ Flex is a specialist finite element analysis package for modelling piezoelectric devices. It
uses an explicit time-domain based approach to solve large finite element problems quickly.
Figure 74 shows two images of models of piezocomposite materials; on the left the material
is in its undisturbed state whereas on the right the material is undergoing its transient
response to an input voltage.
Figure 74 Model of the piezocomposite material in PZ Flex. The left image shows the
material at rest and the right shows the material undergoing mechanical distortion on
application of a voltage due to the piezoelectric property of the ceramic pillars.
An alternative active material considered in the DISSIMILAR project is piezoelectric single

crystals, which are known to have higher electromechanical coupling coefficients and higher
permittivity in comparison to piezoceramics. Figure 75 shows that the single crystal material
(the trace in purple) has a smaller pulse-echo wave train hence temporal resolution and
greater bandwidth than the piezocomposite material (the trace in black). In general, the use
of single crystal material was thought to provide greater resolution, S/N performance and
penetration. However, single crystal materials are more expensive than piezoceramics and
hence a decision was made to not explore their use in the DISSIMILAR project.
Ref: TP AA058J-FR
Version: 01.05
Figure 75 Pulse-echo and bandwidth performance of single crystal (purple) vs. piezoceramic
material.
The primary characteristics of concern to the inspection of austenitic welds are (1) to
decrease the number of cycles in the pulse in order to increase temporal resolution and (2)
increase the bandwidth to increase penetration of attenuative materials by having
appreciable low frequency content. In addition, a further requirement is to reduce ‘cross talk’,
ie energy that leaks from one element to another. Simulation tools exist to predict these
characteristics and optimise them by changing the array materials and geometry. The choice
of the ‘backing material’ plays a crucial role in controlling the level of damping applied to the
resonant active material.
The finer details of materials design and manufacture will not be presented in this document.
In addition to selecting and configuring the active materials, second level specification tasks
for array probes include design of interconnects, cabling and connectors as well as suitable
backing materials/design. These aspects are considered the intellectual property of the probe
manufacturer but it is important to be aware of issues surrounding the manufacture of array
probes when designing the inspection of austenitic welds. The following is a list of key probe
characteristics which must be well conditioned during the second level specification:
1. Element conductance: usually the array is required to be unimodal, ie each element

must be identical to the other. The electrical conductance of each element is a good
indicator of this uniformity as it is highly dependent on element construction (ie the
piezocomposite) and the matching layer thickness.
2. Pulse length: this is specified in the first level specification and is measured to the
20dB drop value from the peak.
3. Bandwidth: this is related to pulse length and is also specified in the first level
specification. The bandwidth is often measured using a Fast Fourier Transform of the
time domain pulse length. The quoted value of bandwidth is calculated as:
[(width @ 6dB drop from peak (in MHz)) / peak frequency (in MHz)] %
4. Cross talk: is specified in the first level specification and refers the signal level that
leaks from one element to another directly.
5. Parasitic echoes: these refer to any internal echoes generated within the array which
is unwanted.
At the end of the second level specification, the manufacture of the array can begin with
confidence that the array will provide the performance stated in the first level specification.
Ref: TP AA058J-FR
Version: 01.05
4.4.3 The TRL-1 prototype for immersion coupled inspection
Figure 76 shows one half of the TRL-1 array manufactured for the DISSIMILAR project: 32x2
configuration, 4mmx6mm pitch and 1.5MHz central (peak) frequency.
Figure 76 A half of the TRL-1 array, as received after manufacture.
Post-manufacture the arrays must undergo a series of tests to (1) ensure array is functional
and (2) meet the requirements of the first level specification. Standard tests include:
1. Element check: each element is checked to see if it is active, establish pulse length
and bandwidth.
2. Calibration block check: standard calibration blocks with SDH targets are used to
study the whole array when focusing and steering the sound field
3. Laser vibrometry: may be used to check the uniformity in displacement exhibited by
each element on the array. Non-uniform vibration would indicate problems in the
piezocomposite and/or electroding.
Figure 77 shows the electrical conductance and capacitance of all the elements on the array,
showing uniformity within 5% of the mean (black line is mean, dashed red lines is +/- 5% of
mean). This indicates that the elements are sufficiently uniform. Additionally the cross talk
was measured to be <-44dB at 1.5MHz, where the specification required it be <40dB (see
Appendix B); hence cross talk performance met the specification. Similarly the parasitic
echoes were found to be within specification (<60dB).
Figure 78 shows results of the laser vibrometry measurements which indicate that the
displacement (both magnitude and phase) of the element shown is uniform over its surface
area, along with the profile across the two lines which similar show sufficient uniformity.
Ref: TP AA058J-FR
Version: 01.05
Figure 77 Electrical conductance and capacitance measurements on the TRL-1 array.
Figure 78 Laser vibrometry measurements showing the magnitude, phase and profile across
the two white lines.
Ref: TP AA058J-FR
Version: 01.05
Figure 79 shows the pulse length and the bandwidth provided by the array (typical). The -
20dB pulse length of the 1.51µs is well within specification (which required it to be <1.69µs).
The bandwidth was measured to be 66%, which was below the 80% required. However,
since the aim of the DISSIMILAR project was to push the manufacturing capabilities, the
value of 66% was deemed higher than what is typically available in the market at the time,
and the ideal goal of 80% remains to be achieved.
Figure 79 The pulse length and bandwidth measured on the TRL-1 array using pulse-echo.
Figure 80 shows the array half setup in immersion for a typical phased array inspection.
Figure 80 The 148mm long half of the TRL-1 array configured for a sector scan on a carbon
steel calibration block containing 1mm SDHs.
Ref: TP AA058J-FR
Version: 01.05
The data is shown in Figure 81, with all the targets aligned as described in Figure 80. The
target in the cursors (which are at an angle of 60°, depth of 38.5mm and horizontal stand off
68mm) is well resolved; however at the higher beam angles (>70°) the beam width is not
sufficiently small to resolve the targets well. Note also the characteristic large dead zone
near the surface due to the pulse-echo setup.
Figure 81 A sector scan data of the functional test described in Figure 80.
TRL-1 is an immersion coupled probe and the required roof / wedge angles were achieved
using a mechanical housing frame, as shown in Figure 82 with all measured parameters.
Figure 82 The final configuration of the TRL-1 probe, ready to be used for inspection.
Ref: TP AA058J-FR
Version: 01.05
4.5 2D-1 array
The final specification of the 2D-1 array is given in Appendix C. SimulUS was used for the
first level specification and PZ Flex was used for the second level specification. The methods
used for arriving at the values for the various parameters of the 2D-1 array was similar to that
described for the TRL-1 in Sections 4.3 and 4.4, and will not be presented in this document.
The 2D-1 array differs from the TRL-1 array in many respects but was constrained by the
128 maximum available channels. Hence the first level specification involved the study of
several configurations:
1. 16x9 with 4 elements at each corner not active

2. 21x6 and 18x7, which only require 126 channels
3. 16x8
4. 21x12, which is termed a 1.5D array configuration
The 2D-1 array was a two-dimensional array (ie 2D), which coupled the sound into the
component using Rexolite as a wedge (rather than using water as for TRL-1) and it was not
implemented in transmit-receive mode, ie it was used in the pulse-echo mode. The 2D-1
array was designed to optimise both the steering and focusing performance required to cover
the full weld of the DISSIMILAR specimen but also to implement skewing of the sound beam,
which is to be able to steer the beam in the plane parallel to the secondary axis of the probe.
The ability to skew the beam has been shown to provide further information on the
orientation of a flaw and allow for full three dimensional manipulation of the sound field.
Figure 83 shows the 2D-1 array on its Rexolite wedge and the metallic mast, through which it
can be attached to the scanning frames.
Figure 83 The 2D-1 array on its Rexolite wedge of 18.5° designed to generate longitudinal
waves in the component.
Ref: TP AA058J-FR
Version: 01.05
4.6 Advanced array configurations for future exploration
Several different array configurations were considered in the DISSIMILAR project, but were
not manufactured and used for inspection. These include annular/segmented arrays,
spiral/fractal arrays, sparse arrays and 1.5D arrays.
The simulated and experimental results for the 2D matrix array demonstrated that it is
possible to produce an array which can skew the beam up to 10°, without introducing
artefacts into the image (see Section 7). Where larger skew angles are required, matrix
arrays with periodic spacing are limited, due to the large number of elements required to
achieve sufficient resolution without forming grating lobes. Even the large array controllers
that are currently on the market with 256 and even 512 channels cannot support these dense
matrix arrays, which in many cases would require upwards of 1,000 elements. Alba
Ultrasound has been developing a series of fractal based array patterns, which overcome
this problem through their aperiodic design. Grating lobes are eliminated, allowing full
volumetric steering with much smaller arrays than could be previously considered. Alba
Ultrasound has constructed a prototype device along a similar specification to the 2D-1
matrix array, which is capable of skewing its beam up to 60° outside the scan plane, using
only 127 elements. For comparison, a 2D matrix design would require 540 elements to
achieve this performance.
To illustrate the difference in capabilities between a standard configuration matrix arrays, ie

the 2D-1 developed for the DISSIMILAR project, and a design based on a fractal
configuration, trials were undertaken to skew the sound beam by up to -60° to +60°. Figure
84 shows the result using the 2D-1 array where the sound field was skewed from -30° to 30°;
the presence of grating artefacts are noted. To compare, an array of the same frequency,
similar number of elements (26 elements in 2D-1 and 27 elements in the fractal array) and
coupled through the same Rexolite wedge was fabricated by Alba Ultrasound. Figure 85
shows the result using the fractal array where the sound field was skewed from -60° to 60°
with the absence of diffracting artefacts.
Figure 84 Result of skewing the sound beam using the 2D-1 array in-excess of 10°, leading
the generation of grating lobes.
Ref: TP AA058J-FR
Version: 01.05
Figure 85 Result of skewing the sound beam using the fractal array designed by Alba
Ultrasound by up to 60° without the generation of grating artefacts.
These results hold promise to be able to fabricate arrays with low element counts (hence
reduce the associated instrumentation costs) which would allow for full 3D control of the
sound field within the component being inspected, opening the possibilities for designing
ultrasonic techniques with greater versatility.
4.7 Specification flow chart and standardisation of probes
The procedure to undertake a first level specification for phased array inspection can be
generalised, independent of array configuration. Starting with the assumption that focusing
will be required, then the near zone length must be greater than the maximum required focus
range, and then the pitch must be suitable for steering range and so on. This procedure can
be visualised in the specification flow chart shown in Figure 86 which captures the steps and
criteria for selecting the key parameters for the TRL probe. The generic design of any
phased array probe consists of iterative loops to optimise key inter-dependent parameters.
In conventional ultrasonic testing, the ability to standardise probes has greatly aided those
who provide inspection services and the generation of industrial application standards. The
use of phased array technology and in particular designing the technique through the use of
flow charts to select the probe will inevitably lead to the generation of a wide variety of
specifications. However, it may still be possible to select a group of probes which would be
prescribed for the majority of inspection tasks. The prescribed probe set will be dictated by
the task; for example, the inspection for (1) root flaws / (2) lack of fusion / (3) cap flaws of
butt welds in carbon steel plates of thickness less than 25mm could be achieved through the
use of a 4MHz linear array probe with 32 elements.
Standardisation efforts should ideally be coordinated by a body such as the British Standards
Institution as part of efforts to generate standards for application of phased array inspection
in industry [15, 16].
Ref: TP AA058J-FR
Version: 01.05
Figure 86 The iterative procedures to follow for setting the parameters of the TRL phased
array probe through modelling the sound field.
Ref: TP AA058J-FR
Version: 01.05
5 Instrumentation
To implement phased array techniques a more sophisticated equivalent of the ultrasonic flaw
detector is required. These instruments, often termed array controllers, are essentially
composed of higher specification versions of the simple flaw detector unit, with the key
difference that several independent pulser-receiver channels are required. This is because
each element of an array is considered an independent transducer or probe and hence must
be operated by its own transmitter-receiver electronics. The consequence is that, in general,
array controllers are at least an order of magnitude more expensive than a standard flaw
detector.
For application to the inspection of thick-section austenitic and dissimilar welds the
bandwidth of the pulser-receiver electronics must be sufficiently low to cover the low
frequency probes that would be required. The pulsing voltage should be 200V or higher to
impart sufficient energy to penetrate the high attenuation weld material. For operation in
metals the time delay resolution, ie the delay between the firing of elements on the arrays,
must be 1ns. Additionally, capabilities such as apodisation (ie the ability to change the
pulsing voltage to different elements) can be beneficial in certain scenarios.
Filters are often applied to the receive circuits of the array controller such that the signals can
be actively processed on reception before storage or display. Band pass filters, ie filters
which allow through frequency content between certain ranges, can often be used to limit the
recorded noise levels. Further active processing can be applied within the electronics to, for
instance, implement fast Fourier transforms (FFTs) to analyse the signals in the frequency
domain rather than the time domain.
The number of independent pulser-receiver channels onboard an array controller impacts on

the phased array techniques that can be developed by limiting the total number of elements
on a probe that can be addressed and the number of elements that can be used to build
focal (or delay) laws. For application to thick-section dissimilar welds, a minimum of 128
independent pulser-receiver channels will be required.
For implementation of full matrix capture techniques (see Section 10.2) the array architecture
needs to be optimised for fast transfer of data from the array controller to the data storage
unit (ie computer). Full matrix capture refers to the recording of data over all possible
transmit-receive combinations of the array; the stored data is then processed by a computer
before being displayed. Hence, for real-time imaging, the data capture and processing must
take place rapidly and increasing the data throughput from the array controller to the
computer to be processed is critical.
The DISSIMILAR project made use of the array controller termed MicroPulse 5PA containing
128 channels (or pulser-receiver units); the specifications of the MicroPulse, manufactured
by Peak NDT of Derby, is given Appendix D. At the time of writing, Peak NDT have
developed and started manufacture of the next generation of array controllers, termed
MicroPulse FMC, which is specifically designed to advance and promote the use of full
matrix capture techniques in industry.
Further detailed discussion of the array controller instrumentation is not included in this
document as the electronic technology in this field is rapidly advancing to enable real-time
imaging capabilities. The reader is referred to the many manufacturers on the market for
detailed information regarding their products and capabilities.
Ref: TP AA058J-FR
Version: 01.05
6 Techniques
The DISSIMILAR project was initiated as an attempt to overcome the considerable difficulties
of inspecting austenitic welds. With the ever increasing development in the fields of
modelling and array ultrasonics, coupled with the quantification method described in Section
3, it was thought that the tools had matured enough to address the fundamental problems of
sound propagation in the inhomogeneous anisotropic austenitic materials. The aim in the
DISSIMILAR project was then to develop a new kind of inspection technique and procedures
tailored for application to the specific case of austenitic materials.
Two complementary routes were considered for (1) overcoming the distortion to improve
sensitivity and (2) correct for the positioning errors. Only the route to improve the sensitivity
was explored in the DISSIMILAR project.
Route 1: Translation tables
In this approach, the distortion induced by the anisotropic weld is first evaluated and then the
positioning is corrected. The modification is specified in the interpretation procedure using
translation tables or lookup tables. The steps required to implement this route are:
Section a reference weld specimen

Map the austenitic weld structure incorporating the crystallographic information
Input the weld map into a model
Perform simulated scans using point reflectors in all positions in the weld
Correct for error in the interpretation procedures
The translation tables are generated using a point reflector introduced into the virtual weld
and inspections are then simulated at the different beam angles used in the actual
inspection; the plotting position of the target is then recorded in a table (assuming a nominal
value for the velocity). The process is repeated with the point reflector in a different position
until the full weld volume has been covered by the reflector at a suitable resolution. This
approach is simulation intensive and is illustrated in Figure 87.
Route 2: Adapted delay laws
This approach is similar to that taken in Route 1 but, instead of using the knowledge of
distortion in the interpretation procedure, it is used to adapt the focal laws used for the
inspection. Note that this technique only improves the inspection sensitivity in its limited
form of application explored in the DISSIMILAR project.
Section a reference weld specimen

Map the austenitic weld structure incorporating the crystallographic information
Input the weld map into a model
Generate adapted delay laws which compensate for distortion
Specify in scanning procedure for inspection
Since the computation times involved in undertaking simulations can be long, only Route 2
was investigated as part of the DISSIMILAR project and the ideas of Route 1 was used to
undertake a limited validation of the quantification method and modelling.
Ref: TP AA058J-FR
Version: 01.05
Figure 87 Generation of the translation tables through the simulation of the inspection for
different beam angles, probe positions and target positions.
In this section, the Route 2 method (termed adapted delay laws) to generate a new
technique is described along with the baseline techniques. The baseline techniques
presented in this document are representative of the vast majority of ultrasonic techniques
used for the inspection of austenitic welds. The DISSIMILAR specimen was inspected using
all the baseline techniques and their performance is presented in Section 7 in comparison to
each other and the sectioning results. The adapted delay laws technique was only applied for
the inspection of Flaw 3 and Flaw 4, the results are again presented in Section 7. The
timescales and complexities of development only allowed a limited investigation of this new
technique in the DISSIMILAR project.
6.2 Adapted delay laws (ADL) technique
In this technique the quantified weld is input to the model, which is then used to generate the
adapted delay laws (ADL). Delay laws refer to the firing times of the elements in an array,
such that the subsequent sound field (or the propagating sound wave generated in the
medium) behaves in the way required. In phased array techniques the two major reasons for
wanting to control the sound field are (1) to direct the wave in a particular direction (ie beam
angle) and (2) to focus the sound energy at a particular position in order to increase the
returned echo amplitude from an indication at that position. Hence the aim of the delay laws
is to ensure that the waves emanating from each element of the array arrive at the correct
time at a position for constructive interference to take place. The delay laws can be
computed from knowledge of the path that the sound will take from each element on the
array, which is a straight line (obeying Snell’s law at boundaries where there is a change in
acoustic impedance). In an austenitic weld there are numerous boundaries (ie grain
boundaries) and additionally the velocity of the sound changes across those boundaries,
hence the simplistic geometric assumptions are no longer valid such that delay laws can no
longer be calculated. In practice, when implementing phased array inspection (as in Section
6.2.3) the effect of the weld is ignored and delay laws are calculated, assuming it is a typical
isotropic weld (such as a carbon steel weld). This then gives rise to the issues illustrated in
Appendix A.
Ref: TP AA058J-FR
Version: 01.05
The ADL method implemented in the DISSIMILAR project is applied primarily with the aim of
improving the sensitivity of the inspection. The material induced error in target position is
corrected for implicitly in the forward transmission, but the software used to plot the returning
echoes still assume a single velocity and straight line paths; hence the error due to incorrect
plotting remains uncorrected when using the ADL technique. For a holistic solution that both
improves sensitivity (overcoming the distortion on transmission) and reduces the error in
positioning, both the ADL concept and the translation tables must be implemented together.
Before the ADL can be generated, the information provided by the weld quantification
(Section 3) must be input to a model. A range of models exist which can be used, including
semi-analytical models such as CIVA and finite element models such as PZ Flex. The work
done in the DISSIMILAR project made use of the CIVA platform, primarily because it was
designed exclusively for the purposes of simulating ultrasonic inspections, whereas PZ Flex
and other models are more generic where it is difficult to simulate complex ultrasonic
inspection scenarios (such as the use of phased arrays). In this section the steps to generate
an adapted delay law is explained, along with the initial evidence showing its potential to
improve inspection. Results of ADL being used for inspection of Flaw 3 and Flaw 4 are
presented in Section 7 and were used to assess any benefits to be derived.
6.2.1 Inputting the quantified weld to the model
The weld map generated using the methods of Section 2 must be input to the CIVA model.
Figure 88 shows the root region of an austenitic weld showing the buttering on the ferritic
steel oriented horizontally. The important task in translating the weld map into the model is to
identify closed regions with clear boundaries, which was done with a combination of
digitisation and manually selecting the boundaries of interest. Subsequently, the weld map of
Figure 88 was transformed to a weld map suitable for the model, as shown in Figure 89.
There are 89 distinct regions and 1276 distinct boundaries in the model map of Figure 89.
The orientation values relevant to each region are then input explicitly to the model along
with the cubic stiffness constants. The weld cross section is then extruded to create the three
dimensional component, ie it assumes that the microstructure remains constant along the
welding direction (Section 3.6).
Figure 88 The root region of an austenitic weld showing the major regions in key colours.
Ref: TP AA058J-FR
Version: 01.05
Figure 89 The EBSD weld map translated into a model map with closed regions.
6.2.2 Generating the ADL
Once the model of the weld is ready, simulated scanning can be undertaken in the CIVA
platform. The generation of ADL is based on time reversal concepts, where the time taken for
echoes to return from a target is converted into the delay laws. Essentially, ADL are not
calculated (as in traditional delay laws using algorithms in the delay law calculator) but
generated using simulation results.
Firstly, a point target (ie a SDH) is introduced at a position in the weld where the sound
energy is to be maximised (ie focused). Then, each element is fired such that a widely
divergent sound field (ie a point source) emanates from the element and enters the weld. In
the weld the propagating wavefront will be distorted and upon incidence on the point target
will return along the same path and the signal is received on the same element. Hence, the
pulse-echo data from the element is collected. The time to arrival of that echo depends on
the specific path that the wave travelled in the weld. The same process is then repeated for
all the elements on the array.
When completed, the travel times to and back from the target is known for all elements on
the array. Hence now, it is possible to fire all the elements such that the waves from all the
elements will arrive at the position of the target at the same time, and in phase. This then is
the generated ADL to maximise the incident energy (ie increase sensitivity) to a target at the
position considered.
Generating the ADL does not require the introduction of SDHs in the actual component or in
reference welds; the ability to quantify the weld using EBSD to such high resolutions and the
use of models such as CIVA has allowed the possibility to undertaken what would have been
a very expensive route much more cost effectively through simulations.
6.2.3 Example of inspecting Flaw 3 using ADL
Figure 90 shows an inspection scenario where a 2MHz 32 element linear array is setup to
inspect Flaw 3 in the DISSIMILAR weld; the beam angle incidence is specular on the flaw.
Ref: TP AA058J-FR
Version: 01.05
Figure 90 The inspection scenario for a specular incidence inspection of Flaw 3.
Figure 91 shows the linear array probe generating the 46.5° beam focused at the depth of
Flaw 3 by using a delay lay calculated assuming that no weld exists.
Figure 91 Integrity of the sound beam focused at the depth of the flaw with no weld.
Figure 92 shows the same delay law applied to generate the beam when the anisotropic
weld is present and the distortion of the beam is clear. The loss in beam strength at the
range of the flaw in comparison to the case when the weld was absent is greater than 20dB.
Ref: TP AA058J-FR
Version: 01.05
Figure 92 The distortion of the sound beam due to the weld when the isotropic delay law is
applied.
Figure 92 illustrates the difficulties inherent to the inspection of austenitic welds in a similar
way to Appendix A where experimental data was presented. Figure 93 shows the A-scan
signal from Flaw 3 when using the calculated delay law and Figure 94 shows the signal when
the generated ADL is used. The maximum amplitude is on the flaw centre when using the
ADL and the absolute gain in signal strength in comparison to when the isotropic law was
used (Figure 6.7) is around 2dB. This result indicates that, theoretically, it is possible to
improve the inspection by being able to project energy better into the required inspection
regions through the use of ADL. Note the shape of the ADL in Figure 94 in comparison to
that of the calculated shape in Figure 93, which indicates that the path taken by the waves is
not based on simple geometric relationships.
In the DISSIMILAR project, ADL were generated for the inspection of Flaw 3 using a linear
phased array probe and for inspection of Flaw 4 using a half of the TRL-1 array. Flaw 3 was
inspected with the beam arriving specular at the flaw through the weld such that the well
validated Kirchhoff interaction method was used. However, the incident beam on Flaw 4 was
not specular and the validity of the theory (Geometric Theory of Diffraction) used to evaluate
the interaction has not been fully established. The performance of the ADL in comparison to
experimental data and detailed information of the inspection setup is presented in Section 7.
Additionally, the ADL for Flaw 3 was generated using version 9.2b of CIVA (32-bit) whereas
the ADL for Flaw 4 was generated using version 10.0 (64-bit); the implications of the change
in version will also be discussed in Section 7.
Ref: TP AA058J-FR
Version: 01.05
Figure 93 The A-scan signal from Flaw 3 when using the calculated isotropic delay law.
Figure 94 The A-scan signal from Flaw 3 when using the generated adapted delay law.
6.3 Baseline techniques
The aim of all the baseline techniques was to inspect 100% of the DISSIMILAR weld and
report indications with (1) position, (2) orientation and (3) through-wall size. It was not an
explicit requirement to characterise the indications into typical weld flaw types, however
knowledge of their nature was implicitly required. There were five baseline inspections, two
of which made use of the probes developed in the DISSIMILAR project: TRL-1 and 2D-1.
All scans were performed from both sides of the weld with the probes placed on the top
surface (OD surface); in the case of the automated baseline (Section 6.3.2), the inspections
Ref: TP AA058J-FR
Version: 01.05
were performed from the bottom surface also, the ID surface. Note that the ID surface is
where the root of the V-prep weld in the DISSIMILAR specimen has been deposited.
6.3.1 Manual conventional technique
The SONATEST Powerscan 400 flaw detector was used which provided a 400V pulser along
with gain increments in 0.5dB. The probes used were conventional probes designed for
inspection of austenitic welds:
1. 0° 2MHz twin-crystal 20mm diameter compression probe

2. 45° 2MHz twin-crystal angle compression probe (Krautkrämer type VRY)
Calibration of time base was performed in accordance with a draft standard for the inspection
of austenitic welds [15]. There were several exceptions to the recommendations of the draft
document [15] which are listed below:
1. A reference block (with targets in a representative weld) was not used for generation
of the distance amplitude correction (DAC) curves
2. Sensitivity was based on 5-10% full screen height (FSH) grain interference (grass
noise)
3. The 20dB beam spread was not established for all the beams as a reference block
with the weld was not available
4. The max amplitude technique was used for sizing all indications
Inspection was performed by a Level III ultrasonic operator experienced in austenitic welds.
6.3.2 Automated (encoded) conventional technique
The automated technique was performed using the MIPS/GUIDE system developed by
British Energy Generation Limited (BE). Details of the instruments, equipment and probes
used in this baseline are available in the inspection report presented in Appendix D. This
baseline represents the current best practice inspection being implemented in the nuclear
power plants of the UK.
6.3.3 Linear phased array technique
This is the first of three phased array technique which made use of a 32 element 2MHz linear
array probe with a pitch of 1.5mm and width 20mm. The MicroPulse (Section 5) was used
with the software ArrayGen (version B1008) to collect, process and display the data; the data
was displayed and interpretation was done in the sector scan format. The probe was
mounted on a Perspex wedge of 20° with a wedge thickness (as defined in ArrayGen) of
16.5mm.
The technique was conducted in accordance, wherever possible, with the draft standard for
inspection of austenitic welds [15] and with the draft standard for inspection using phased
arrays [16]. As for the manual technique (Section 6.3.1) an exception was the absence of a
DAC generated in a reference weld, which was not available. Time base calibration was
performed for all the angles in the sector using a 304 stainless steel calibration block
containing 3mm SDHs with a measured nominal velocity of 5750m/s. Sensitivity, as in the
manual technique, was taken with reference to the grass noise level. However, sizing was
Ref: TP AA058J-FR
Version: 01.05
performed using tip diffracted echoes, whenever these could be established with some
confidence.
The array was used to focus the sound energy to the depth of the target and each flaw was
investigated individually over the beam angles available in the sector.
6.3.4 TRL-1 phased array technique
The inspection using TRL-1 was done using the BE MIPS/GUIDE system configured to
implement phased array inspection. The array probe was programmed to generate beams at
a range of angles (as in a sector scan). The data was collected using MIPS (version
1.23.T8), then processed and interpreted using GUIDE (version 1.14.T1). A carbon steel
calibration block with 3mm SDHs was used for the calibration of time base and sensitivity
was (as for all baseline techniques) set to the grass noise level. As in the case of the linear
phased array technique, tip diffraction methods were used for sizing whenever possible.
The array was again used to focus the sound energy to the depth of the target and each flaw
was investigated individually.
6.3.5 2D-1 phased array technique
As for TRL-1, the inspection using 2D-1 was undertaken using the MIPS/GUIDE system.
Similarly, all other aspects regarding calibration, sensitivity and approach was similar to that
for the TRL-1 inspection described in Section 6.3.4, except the use of skewing. Skewing is
when the sound beam is steered not just in the plane of the main probe axis, but on the
secondary plane also. The 2D-1 probe was used with a number of skew angles in the vicinity
of flaws to investigate improved methods of establishing and dealing with flaw orientation.
Ref: TP AA058J-FR
Version: 01.05
7 Performance
This section discusses the performance of all the techniques to detect and size the flaws that
were introduced into the specimen. The performance of the baseline inspection techniques is
discussed in Section 7.2, followed by the performance of the ADL technique explored in the
project in Section 7.3 and efforts towards validation of the models in Section 7.4.
7.2 Baseline inspections
The primary aim of the project was to establish the positioning and sizing capabilities of
phased array ultrasonic techniques against other established ultrasonic techniques; all the
techniques explored in the project are described in Section 6.3. All the baseline techniques
were used to inspect the DISSIMILAR specimen (see Figure 5) to detect, position and size
the implanted flaws. Flaws 1, 2, 3 and 4 were then sectioned to confirm against the nominal
positions and sizes; flaws 5, 6, 7 and 8 were retained for future experimental demonstrations
and trials to compare against the work done in the DISSIMILAR project.
The general requirements of ultrasonic inspection are to establish:
1. The type of flaw

2. The position of an aspect (eg the top tip) of a crack from a reference (eg top surface)
3. Size; both through thickness and lateral
The operators who undertook the baseline inspections were not required to establish the flaw
type; it was assumed that the range of flaws introduced could be quite readily characterised
through the data obtained. The positions and sizes reported using the baseline inspection
techniques were compared to the nominal sizes, which were checked through sectioning.
The performance of each technique for Flaws 2 to 8 is quantified in charts where the
through-wall extent bars show the flaw height and measured distance from the top surface of
the specimen to the top edge of the flaw, and the (lateral) flaw length bars give the start
position and measured length of the flaw along the welding line. The inspection data from
using the various techniques are presented in this section, except for those from the
conventional automated scans which are given in Appendix E.
In the figures that follow showing the through-wall and length measurements provided by the
different baseline techniques, the key to the data labels are as follows:
Nominal: as implanted, confirmed by sectioning for Flaws 2, 3 and 4;

PA 2D (AA): 2D-1 phased array technique (see Section 6.3.5) where AA is the side of
the weld centreline where the probe was placed - either SS for stainless steel side or
CS for carbon steel side;
TRL (AA): TRL-1 phased array technique (see Section 6.3.4) where AA is the side of
PA (AA): linear phased array technique (see Section 6.3.3) where AA is the side of
Conv. Ma (XX): conventional manual technique (see Section 6.3.1) where XX is the
beam angle of the probe used to collect the data;
Ref: TP AA058J-FR
Version: 01.05
Conv. Au (BB AA): conventional automated technique (see Section 6.3.2 and
Appendix E for full inspection report) where (1) BB indicates whether the scanning
was done on the ID for inside diameter surface or OD for outside diameter surface
and (2) AA is the side of the weld centreline where the probe was placed - either SS
for stainless steel side or CS for carbon steel side;
Conv. Au (All): indicates the closest measurement found from all the probes that were
used in this baseline inspection.
Note that only the conventional automated technique was performed from both inspection
surfaces, termed ID and OD, and all the other techniques were performed from the OD
surface only, which is the top surface opposite to the weld root.
7.2.1 Flaw 2
Figure 95 shows the performance of all the techniques for Flaw 2. The greatest positional
error through-wall was when using the TRL-1 technique, which positioned the top edge of the
flaw 5mm below the surface (error of -5mm). The conventional automated technique overall
had a lesser error in through-wall position (-1mm) but the greatest error in through-wall size
(+3mm). None of the techniques were able to measure both the flaw position and size to
within +/- 1mm.
Figure 96 shows the results of the conventional manual technique for plotting Flaw 2 when
using the 45, 60 and 70˚ beams; the technique used the max amplitude sizing method. There
were substantial changes in both the plotted position and evaluated through-wall size when
using the three different beams.
Note that all three beams have positioned the flaw on the weld side of the buttering, whereas
the sectioning (see Figure 97) shows that the flaw was introduced on the parent ferritic side
of the buttering. However, there will have been some measurement error (estimated to be
around 2mm) when measuring from the centreline and in addition changes in the actual
beam angles / probe index points. The primary parameter which would have affected the
plotted position of Flaw 2 would likely have been the assumed velocity (6042 or 5626m/s
depending on whether inspection was from ferritic or stainless steel sides, respectively).
Figure 95 Graphs showing the comparison of the position, through-wall and length
measurements made by the baseline techniques for Flaw 2. Refer to key on page 95.
Ref: TP AA058J-FR
Version: 01.05
Figure 95 shows that the lateral flaw length measured by all the techniques, except the TRL-
1, was accurate to the nominal measured value to within 2mm. However, the flaw start
position was measured incorrectly by all the techniques. The TRL-1 technique measured the
flaw to be 7mm longer than nominal and starting earlier with respect to the datum, ie starting
at 77mm whereas the nominal flaw start was measured close to 85mm.
In general, over all the other flaws also, the performance of the TRL-1 technique for lateral
flaw sizing was found to be poorer than the other techniques. It is known that twin element
probe configurations (such as the TRL probe developed in DISSIMILAR) have an optimum
working volume which is constrained by geometric parameters, in particular the roof angle.
The use of two dimensional arrays on each half of the twin probe allows for some control and
widening of this working volume; however, each half of the TRL probe developed had a 32 x
2 configuration and it was not possible to greatly influence the working volume using only 2
elements along the lateral axis. Future designs of the TRL probe should increase the number
of elements along the width of the array (hence also increase the total number of channels in
the instrument) to improve the lateral sizing ability of the TRL-1 type techniques.
Ref: TP AA058J-FR
Version: 01.05
Figure 96 Plotting of Flaw 2 using the conventional manual technique; the 45˚ beam (top),
the 60˚ beam (middle) and the 70˚ beam (bottom) were used in the technique.
Ref: TP AA058J-FR
Version: 01.05
Figure 97 Sectioned view of Flaw 2 through the central point along its length.
Inspection of Flaw 2 was accomplished most readily by all the techniques from the stainless
steel side where the sound had to travel through the weld, rather than through the clean
ferritic parent side. With the probe placed on the ferritic carbon steel the flaw was not
oriented well for specular detection, which then had to be detected through tip diffraction
signals. Figure 98 shows the detection and sizing of Flaw 2 using the linear phased array
(PA) technique with the probe placed on the stainless steel side and Figure 99 shows the
same flaw from the ferritic side. In both cases the flaw could be both detected and sized;
however the S/N of the diffracted signals is weaker when the probe is placed on the ferritic
side.
Ref: TP AA058J-FR
Version: 01.05
Figure 98 Detection and sizing of Flaw 2 using the linear phased array (PA) technique with
the probe placed on the stainless steel side of the specimen.
Figure 99 Detection and sizing of Flaw 2 using the linear phased array (PA) technique with
the probe placed on the ferritic side of the specimen.
Figure 100 shows that there were differences in the measured orientation of Flaw 2 using
inspections from either side of the weld centreline. The weld bevel angle to the vertical was
designed to be 15˚ but was measured to be 20.6˚ from the stainless steel side (ie when
sound travelled through the distortive weld metal) and was 14˚ from the carbon steel side.
This implies that even though detection was easier through the weld metal due to favourable

Ref: TP AA058J-FR
Version: 01.05
orientation of the flaw, the actual measurement of orientation appears to be adversely

influenced by the distortion.
Figure 100 Evaluation of the orientation of Flaw 2 using the linear phased array (PA)
technique with the probe on either side of the weld.
Figure 101 shows the data from the TRL-1 technique where Flaw 2 was detected from the
stainless steel side. Note that the S/N is greater than 6dB and the near surface noise is near
absent. In contrast the technique using the 2D probe shows significant surface noise which
can mask the echoes from the flaw; however, by reducing the effective aperture size (ie the
number of elements) it was possible to detect Flaw 2 but with much less S/N performance,
as shown in Figure 102.
The S/N performance of the 2D-1 technique was in general poorer than the TRL-1 technique
but was found to be similar to the conventional manual and linear PA techniques. The
conventional automated technique (see Appendix E) made use of probes with single and
twin radiating elements and the S/N performance was variable. In the conventional
automated scans specific probes were selected as ‘search’ units for various zones of the
weld and other focused twin probes were used for sizing the indications (see Appendix E).

Ref: TP AA058J-FR
Version: 01.05
Figure 101 Flaw 2 detected and sized using the TRL-1 technique from the stainless steel
side.
Figure 102 Flaw 2 detected and sized using the 2D-1 technique from the stainless steel side
(left) and the carbon steel side (right).
The difficulty for finding and characterising Flaw 2 was its proximity to the top surface and its
orientation, which is not favourable to reflect sound from. It was found that all the techniques
were able to detect and size the flaw, with the approach through the weld metal giving the
greater success.

Ref: TP AA058J-FR
Version: 01.05
7.2.2 Flaw 3
Flaw 3, unlike Flaw 2, was ideally oriented on the buttering to weld interface near the root for
near specular approach of the sound beam from the stainless steel side, travelling through
the weld. Figure 103 shows the summary of performance for all the techniques and Figure
104 shows the sectioned Flaw 3 along its mid-length point.
None of the techniques were able to measure all the parameters (through-wall position,
through-wall size, flaw length start and flaw length) accurately. The greatest error (+7mm) in
through-wall size was recorded by the conventional automated technique (from the stainless
steel side) and the greatest error in length wise positioning and size was recorded by the
TRL-1 technique. The poor performance of the TRL-1 technique in lengthwise
characterisation is thought to be due to the same limitations as in the case of Flaw 2.
Figure 105 shows the nominal position of Flaw 3 along with the conventional manual data
using the 45 and 60˚ beams. As for Flaw 2 the plotted position of the flaw varied when using
different beams. Figure 106 shows the inspection data using the linear PA technique when
the full array was used, but without focusing. The incident angle on the flaw indicates that the
flaw is specular to the beam when the beam angle is ~52˚, however the actual incident beam
angle to be perfectly specular should theoretically be 45˚.

Ref: TP AA058J-FR
Version: 01.05

Ref: TP AA058J-FR
Version: 01.05
Figure 105 Inspection of Flaw 3 using the conventional manual technique; nominal (top), 45˚
beam (middle) and 60˚ beam (bottom).
Figure 106 Detection of Flaw 3 using the linear PA technique when the full array was used
without focusing the sound field.

Ref: TP AA058J-FR
Version: 01.05
By contrast, when the beams were focused to depths near that of Flaw 3 using the linear PA
technique, the flaw was less clearly detected, as shown (in the case of focusing to 50mm) in
Figure 107. Additionally, the variability of the flaw echo along the weld direction was severe
enough for the operator to be unable to confidently measure the through-wall size of Flaw 3
using the linear PA technique.
Figure 107 Detection of Flaw 3 using the linear PA technique when the full array was used to
focus the sound field to the depth of 50mm (flaw was at a depth of 70mm).
Figure 108 shows the detection and characterisation of Flaw 3 using the TRL-1 technique,
focusing the sound field to the depth of Flaw 3; the probe was placed on the stainless steel
side with the sound beam travelling through the weld metal. The S/N performance of the TRL
probe is again significantly better than those of the other techniques (note the S/N of the
linear PA technique in Figures 106 and 107). However, the orientation of the flaw was
unfavourable for inspection from the carbon steel side, as shown in Figure 109 where the
TRL-1 technique was used and even with increasing sensitivity (ie reducing S/N quality) it
was not possible to detect any diffracted signals from the region near the edges of Flaw 3.

Ref: TP AA058J-FR
Version: 01.05
Figure 108 Inspection of Flaw 3 using the TRL-1 technique with the probe placed on the
stainless steel side and focusing the sound to the depth of the flaw.
Figure 109 Inspection for Flaw 3 using the TRL-1 technique with the probe placed on the
carbon steel side and focusing the sound to the depth of the flaw.
Figure 110 shows the detection of Flaw 3 using the 2D-1 technique from the stainless steel
side, with poorer S/N performance in comparison to the TRL-1 technique.

Ref: TP AA058J-FR
Version: 01.05
Figure 110 Inspection for Flaw 3 using the 2D-1 technique with the probe placed on the
7.2.3 Flaw 4
Figure 111 shows the summary of the performance of all the techniques when inspecting for
Flaw 4 and Figure 111 shows the sectioned view of Flaw 4 at its mid-length position.

Ref: TP AA058J-FR
Version: 01.05
The best performance in through-wall positioning and sizing was achieved by the 2D-1
technique (to ~1mm) with the probe placed on the stainless steel side (ie through detection of
diffraction signals), as shown in Figure 113. Note also the echoes from the parent stainless
steel near mid-wall thickness which are thought to be inclusions/segregations and the loss of
near surface resolution due to the very strong (saturated) surface echoes.

Ref: TP AA058J-FR
Version: 01.05
Figure 113 Inspection for Flaw 4 using the 2D-1 technique with the probe placed on the
Figure 114 shows the detection and sizing of the same flaw from the carbon steel side with
the sound going through the weld metal and using the TRL-1 technique, which offered the
best S/N performance for the task but its performance in through-wall positioning had an
error in-excess of 10mm.
Figure 114 Inspection for Flaw 4 using the TRL-1 technique with the probe placed on the
carbon steel side and focusing the sound to the depth of the flaw.
The conventional manual technique was also able to detect and size Flaw 4 using the max
amplitude technique. Figure 115 shows the nominal position of Flaw 4, the plotted position
using a 60˚ beam when the probe is placed on the carbon steel side (ie travelling through the
weld metal) and an illustration of the S/N performance when using the 60˚ beam.

Ref: TP AA058J-FR
Version: 01.05
Figure 115 Inspection of Flaw 4 using the conventional manual technique from the carbon
steel side; nominal (top), 60˚ beam (middle) and the S/N performance of the 60˚ beam
(bottom).
The distortion of the sound field and grain noise was severe enough for effective
interpretation to be not possible when using the linear PA technique; Figure 116 illustrates
the difficulty of identifying the flaw signal confidently against the background grain noise level
when using the linear PA technique with the sound traversing through the weld metal. In
general the inspection for Flaw 4 using the various different techniques proved to be difficult
when the sound had to traverse the weld metal and it was only possible to detect the tip
diffracted signals from the stainless steel side with the system gain at high levels.
The length positioning of Flaw 4 was consistent amongst all the techniques, starting ~180mm
from the datum with a difference of -5mm from the nominal measured value. The greatest
error in flaw length sizing was registered by the conventional automated technique from the
carbon steel side (-7mm).

Ref: TP AA058J-FR
Version: 01.05
Figure 116 Inspecting for Flaw 4 using the linear PA technique with the probe on the carbon
steel side, showing the level of back-scattered grain noise.
7.2.4 Flaw 5
The performance of all the techniques for inspection of Flaw 5 is shown in Figure 117. It was
not possible to detect and size Flaw 5 using the 2D-1 technique from either the carbon steel
or stainless steel sides. Flaw 5 was introduced on the interface between the buttering and
the parent carbon steel (see Figure 10).
Despite the unfavourable orientation Flaw 5 was detected and sized from the carbon steel
side by the TRL-1 technique as shown in Figure 118. When using the TRL-1 technique from
the stainless steel side it was difficult to identify the echoes from the flaw in comparison to
other echoes, some of which were geometric, as shown in Figure 119.

Ref: TP AA058J-FR
Version: 01.05
Figure 118 Detection and sizing of Flaw 5 from the carbon steel side using the TRL-1
technique; two cross-section views are shown on the right at 77 and 80mm from the datum.
Figure 119 Detection of Flaw 5 from the stainless steel side using the TRL-1 technique
showing the cross-section view at 82mm from the datum.
Inspection using the linear PA technique was able to detect (but not size) when the sound
had to travel through the weld (see Figure 120) but the operator deemed it possible (with
some confidence) to size Flaw 5 from the carbon steel using diffracted echoes, as shown in
Figure 121.

Ref: TP AA058J-FR
Version: 01.05
Figure 120 Detection of Flaw 5 from the stainless steel side using the linear PA technique.
Figure 121 Detection of diffracted echoes from the tips of Flaw 5 using the linear PA
technique from the carbon steel side.
The positioning and sizing of the conventional manual technique from the stainless steel side
is likely to have been severely influenced by the weld metal. Figure 122 shows the plotting
performance when using the 60˚ beam from the stainless steel side. Flaw 5 was difficult for
all the techniques but surprisingly the length sizing error of the TRL-1 technique from the
carbon steel side was far worse (+9mm) than when the sound had to travel through the weld
(error of +2mm in the length sizing compared to nominal). However the through-wall size of
Flaw 5 measured by the TRL-1 technique from the stainless steel side was 24mm, an error
of +14mm. This indicates that a scenario where an unfavourably oriented flaw has to be
detected through the weld can present significant problems for ultrasonic techniques.

Ref: TP AA058J-FR
Version: 01.05
Figure 122 The detection and sizing of Flaw 5 using the conventional manual technique from
the stainless steel side; nominal (top), 60˚ beam (middle) and the S/N performance of the 60˚
beam (bottom).
7.2.5 Flaw 6
Figure 123 summarises the performance of the techniques for inspection to detect and size
Flaw 6, which is similarly oriented and positioned to Flaw 4 but is half the size in the through-
wall direction. The conventional automated inspections oversized the flaw height by 5mm
and the TRL-1 technique again substantially undersized the length of the flaw (error of -
10mm to nominal). The best performance was achieved by the 45˚ beam of the conventional
manual technique from the carbon steel side (ie traversing the weld) with good S/N and, in
the through-wall direction, positioning the flaw with an error of +1mm and sizing accurately as
5mm, as shown in Figure 124.

Ref: TP AA058J-FR
Version: 01.05
the carbon steel side; nominal (top), 45˚ beam (middle) and the S/N performance at the 45˚
beam (bottom).

Ref: TP AA058J-FR
Version: 01.05
The 2D-1 technique was able to detect and size the flaw, but with poor S/N performance, as
shown in Figure 125. In contrast, the S/N performance of the linear PA technique from the
stainless steel side was much better, as shown in Figure 126.
Figure 125 Detection and sizing of Flaw 6 using the 2D-1 technique from the stainless steel
side.
Figure 126 Detection and sizing of Flaw 6 using the linear PA technique from the stainless
steel side through the detection of diffracted signals from the flaw tips.
The TRL-1 technique was also able to detect the diffracted signals from the flaw tips and
thereby undertake sizing, as shown in Figure 127.

Ref: TP AA058J-FR
Version: 01.05
Figure 127 Detection and sizing of Flaw 6 using tip diffracted signals using the TRL-1
technique from the stainless steel side.
7.2.6 Flaw 7
The detection and sizing performance of all the techniques in inspection for Flaw 7 is shown
in Figure 128. Flaw 7 was introduced as a rough centreline crack at mid-thickness of the
weld (see Figure 12). All the techniques except the linear PA technique could find and size
Flaw 7; the linear PA technique could only detect the flaw, as shown in Figure 129.

Ref: TP AA058J-FR
Version: 01.05
Figure 129 Detection of Flaw 7 using the linear PA technique from the stainless steel side.
Flaw 7 was also found and sized by the conventional manual technique from the stainless
steel side using 45, 60 and 70˚ beams, as shown in Figure 130. Flaw 7 was readily detected
by all the techniques when travelling through the weld to stainless steel parent interface.
However, the flaw could not be detected when travelling through the buttering layer (ie from
the carbon steel side). This is illustrated by the data from the 2D-1 technique, as shown in
Figure 131, where the flaw is clearly identified from the stainless steel side but could not be
found from the carbon steel side.

Ref: TP AA058J-FR
Version: 01.05
the stainless steel side; nominal (1), 45˚ beam (2), 60˚ beam (3) and 70˚ beam (4).
Figure 131 Detection of Flaw 7 by the 2D-1 technique from the stainless steel side (left) and
from the carbon steel side (right).

Ref: TP AA058J-FR
Version: 01.05
The best S/N performance was again achieved by the TRL-1 technique, as shown in Figure
132. In general, Flaw 7 did not present a challenge to any of the techniques as long as the
beams did not have to traverse the buttering layer. The rough nature of the flaw (ie the
presence of flaw facets at different orientations) presented an advantage for detection but
made the sizing of the flaw a little more difficult as the extreme facets had to be selected for
the sizing.
Figure 132 Detection and sizing of Flaw 7 using the TRL-1 technique from the stainless steel
side.
7.2.7 Flaw 8
The detection and sizing performance of all the techniques in inspection for Flaw 8 is shown
in Figure 133. Note that the TRL-1 technique could not be evaluated as the flaw proximity to
the edge of the specimen in comparison to the width of the probe was not well conditioned
for the TRL-1 technique; note that the need to remove material for the reference/validation
block (see Figure 178 in Section 7.4) and for EBSD led to this compromise. Flaw 8 (see
Figure 13) represents a smooth root lack of fusion 3mm high and breaking the back wall of
the specimen; it was generated as an EDM notch. The presence of both the buttering layer
and the weld metal, and the small size of the flaw, makes its detection and sizing difficult.
However, all the techniques were able to detect the flaw from the stainless steel side and
again the buttering proved to be an obstacle preventing detection from the carbon steel side.
This is illustrated in the case of the 2D-1 technique in Figure 134.

Ref: TP AA058J-FR
Version: 01.05
Figure 134 Detection of Flaw 8 (shown with a white arrow) from the stainless (left) and
carbon (right) steel sides.
Similarly Flaw 8 was quite readily detected from the stainless steel side by the linear PA
technique but could not be sized, as shown in Figure 135. Only the conventional techniques
(both and manual and automated) were able to size the through-wall dimension of the flaw,
reporting 3mm (error of 0mm) in the case of the manual and 5mm in the case of the
automated techniques. Note however that typical conventional procedures specify a size
range (eg <3mm) related to the minimum sizable capability of the technique and hence the
values reported by the techniques represent that minimum sizing capability. In the case of all
the phased array techniques, sizing was only performed when clearly identifiable diffraction
echoes were detected or the operator felt confident that the beam-flaw interaction was at or
close to the specular regime.

Ref: TP AA058J-FR
Version: 01.05
Figure 135 Detection of Flaw 8 from the stainless steel side by the linear PA technique.
7.2.8 General discussion
In summary, the conventional manual technique was able to detect all the flaws using the
grain interference method to set the sensitivity; the plotting and sizing results however varied
from probe to probe. A compression 0˚ probe was used to assess the attenuation on the
specimen by setting the back wall signal to 80% of full screen height (FSH); the gain required
in the parent stainless steel was 45.5dB, that of the parent carbon steel was 51dB and that of
the weld was 74dB. This represents a marked difference between the three regions of the
weld and it is known that the properties of the weld changes with sound propagation
directions (anisotropy), affecting beam angle differentially.
The linear PA technique illustrated the ‘toughness’ of the weld to allow effective sound
propagation, in particular when the sound had to travel through the weld metal or the
buttering. Through-wall sizing errors were up to 5mm and diffraction signals could be
detected effectively only when the flaws were on the fusion boundaries where the sound did
not have to traverse the distortive buttering or weld media. A significant dead zone due to the
surface was evident.
In general, the best S/N performance was achieved by the TRL-1 technique (see Appendix A
for an additional example of using TRL probes) and the worst using the 2D-1 technique.
Additionally, the TRL-1 technique was better at penetrating the buttering layer than the other
techniques but suffered from the worst length sizing capabilities of all the techniques.
The 2D-1 technique was implemented with skewed beams also to explore capabilities to
determine any skew orientation to the detected flaws. The array configuration (18 x 7) was
specified with enough elements along the secondary axis to be able to steer the sound beam
by up to +/- 10˚ in the secondary plane. The secondary plane is parallel to the welding
direction and any flaws which are not perfectly in this plane, ie skewed, would be more
sensitive to an appropriately skewed beam – that is, a skewed beam should increase the
likelihood of detecting skewed flaws.

Ref: TP AA058J-FR
Version: 01.05
Figure 136 shows data on using skewed beams for inspecting Flaw 3. Comparing the signal
strength from the flaw when the beam is skewed -4˚ and +4˚ to when the beam was not
skewed (ie 0˚ beam), the echo strength from the flaw when the beam was +4˚ is reduced
implying that the flaw is tending away from being specular. Hence the preferred skew of the
flaw can be inferred from comparing the skewed data.
Figure 136 Data showing the signal strength from Flaw 3 when the beam is skewed +/- 4˚
from nominal 0˚ beam.
Investigations on Flaw 4, as shown in Figure 137, reveal similar sensitivity to the incident
beam, ie sensitivity to skew, which indicates the orientation of the flaw itself. A negatively
skewed beam (-7˚) leads to a 3dB drop in maximum diffracted signal amplitude whereas only
a 1dB drop when the beam is skewed positively to +7˚.
Data from using skewed beams to inspect Flaw 7, the rough centreline crack, implies that
flaws need not have preferential orientation. Figure 138 shows that both +/- 4˚ beams lead to
similar reflected amplitudes (to within 1dB) to when the beam was not skewed. Even when
the beams are skewed to +/- 8˚ the amplitudes are still within 3dB of that at 0° skew. This is
consistent with the nature of rough cracks which remain reflective due to their faceted nature.
The study showed that the use of 2D arrays allows to infer any tilt or orientation to the flaws
but measuring the actual tilt will require the system to undertake scans with several beams
(similar to a sector scan but over the secondary plane). The specification of the array will
determine the limits to which the beams can be skewed and it is important that these limits
are not exceeded, as the break down in the sound field characteristics could lead to
unforeseen phenomena (primarily through the generation of side energy lobes). Additionally,
skewed beams are more likely to be useful for detecting smooth flaws (which are more
sensitivity to incident beam directions) than rough faceted flaws.

Ref: TP AA058J-FR
Version: 01.05
Figure 137 Data showing signal strength from Flaw 4 when the beam is skewed by +/- 7˚
and +11˚ from nominal 0˚ beam.
Figure 138 Data showing signal strength from Flaw 7 when the beam is skewed by +/- 4˚
and +/- 8˚ from nominal 0˚ beam.
The data from the baseline inspections show that the use of phased array techniques does
not lead to any appreciably improvement in the sizing capability. The work done has
highlighted the areas where the use of phased array techniques could provide significant

Ref: TP AA058J-FR
Version: 01.05
logistical advantages in comparison to both the manual and automated conventional

techniques.
A key parameter that was measured as part of the baseline inspections was the total time
taken to complete the inspection, from when the specimen is presented to when the
inspection report is submitted. The conventional manual inspection of all the flaws in the
specimen took one week, with 5 hours to set up and 30 hours for scanning and
interpretation. Four probes were used, three for the actual scanning. The phased array
manual inspection (the linear PA technique) took 2 hours for the setup (including calibration
of the one linear array probe) and 10 hours for scanning and interpretation.
The conventional automated inspection (see Appendix E) is estimated to have taken a week
including setup, scanning and interpretation. The technique made use of many probes (more
than five), but the inspection was done from both the inside and outside (top) surfaces of the
specimen, whereas all the other inspections were done from only one (the top) surface.
The TRL-1 technique made use of only one probe. The setup time was estimated to be 0.5
hour per flaw (ie 4 hours in total) and 15 hours for scanning and interpretation - the
interpretation was estimated to be 70% of this time. The 2D-1 technique involved the
exploration of various inspection concepts and hence was not counted towards evaluating
inspection times. Hence, in general, the phased array techniques (linear PA and TRL-1) were
less than 50% of the time taken by the conventional techniques. Once inspections are set up,
routine inspections should take even less time to implement using mechanical systems.
Both the TRL-1 and 2D-1 techniques can be automated but only one probe needs to be
deployed, unlike in the case of the conventional automated technique (see Appendix E). The
TRL and 2D probes developed in the project are able to address the full volume of the weld,
whereas a number of specific probes were required to interrogate the full weld volume in the
conventional automated technique. Hence the total time to complete the inspection in an
industrial setting using the phased array techniques should be much less than the
conventional technique, which requires changing several probes from the mechanical
deployment system.
Using a fewer number of probes will also allow the mechanical deployment system to be
simplified, reducing the use of large cumbersome probe pans containing several probes
generating a range of discrete beam angles. This should also allow designers to minimise
the chances of mechanical failures leading to introduction of debris into containment areas.
Another key logistical advantage of phased array techniques (and other automated data
collection techniques) is that they inherently provide digital data that can be archived. The
data representation of phased array techniques (whether the sector scans of the linear PA
technique or the processed representation provided by GUIDE) allow for more intuitive data
interpretation by the human operator and also provide future possibilities for automated
interpretation of data.
The logistical advantages described above are secondary to the measurement capabilities of
the phased array techniques but they come to the fore when they make the techniques
attractive for industry through their ability to reduce implementation costs.

Ref: TP AA058J-FR
Version: 01.05
7.3 Adapted delay laws (ADL) technique
The concept of generating delay laws through the use of models, as described in Section
6.2, was developed in the DISSIMILAR project as a possible route to overcome the
degradation in the inspection quality due to the distortion of the sound by the austenitic weld
[16, 17 and 18]. The results of generating the ADLs using the CIVA model is presented in
Section 7.3.1 and experimental results of using the ADLs to undertake scanning of the
specimen is presented in Section 7.3.2. Additionally, initial results of using a finite element
package (PZ Flex) to generate the ADLs is presented and results from attempts to generate
the ADLs experimentally (rather than through models) is presented.
7.3.1 Simulated
The ADLs were developed as described in Section 6.2. Figure 139 shows the inspection
scenario where the probe was scanned over the flaw with a total of seven positions (P1 to
P13) incident on the flaw, assuming straight geometric path of the beam. The beam angle
was set to be 46.5˚ which was perfectly specular to the simulated Flaw 3 at the position P7
when the beam is incident on the central point of the flaw area.
Figure 139 Simulated scanning of Flaw 3 using a linear array probe with beams from seven
discrete positions being incident on the flaw.
Figure 140 shows the simulated echo from Flaw 3 when the elements were fired individually
in order to generate the ADL at one of the seven positions. Figure 140 shows that the
greatest received echo strength was on elements 15 to 20, implying that the lower elements,
which would have been incident on the weld fusion interface at oblique angles, appear to
lose energy possibly through reflection at that boundary. The received amplitude is not
relevant for the generation of the ADL (see Section 6.2) as only the time-of-flight is used to
‘reverse’ the scenario such that all the element wavelets arrive at the target at the same time
and in-phase for constructive interference to take place.

Ref: TP AA058J-FR
Version: 01.05
Figure 140 The received echo on element 15 (of the 20 element array aperture) showing
that the sound from elements 1 to 15 are weaker relative to the elements 15 to 20.
Hence using the simulated times-of-flight for echoes generated at the simulated Flaw 3, an
ADL was generated for each of the seven positions and the delay law curves for each
position is presented in Figure 141.
Figure 141 ADLs for each of the seven positions forming the simulated scan over Flaw 3.
The degree of change in the profile of the delay law curve and the changes in delay
amplitudes (measured in Nano seconds, ns) is indicative of the anisotropy of the weld
medium through changes in the sound velocity. Whereas the isotropic delay law curve is
near symmetric this is not the case with the ADLs. The simulated echo from the flaw was
then generated when using the isotropic and each of the ADLs separately; the results are
presented in Figure 142. In all cases the echo due to the ADLs (fired from their
corresponding positions) are larger than that due to the isotropic delay law. Note also the
changes in received time at several of the positions between when the adapted and isotropic
delay laws were used.

Ref: TP AA058J-FR
Version: 01.05
Figure 142 Simulated echoes received from Flaw 3 when using each of the seven ADLs
from their corresponding position in comparison to the isotropic delay law.
A further result using the same array as shown in Figure 139 but using 32 elements (instead
of 20 elements as above) and with the array placed in position P7 is presented in Figure 143,
showing a significant increase in the received amplitude from the flaw when using the
adapted delay law. The received amplitude from each of the positions was then plotted to
generate the echo-dynamic over Flaw 3; additionally, the echo-dynamic of Flaw 3 assuming
that the weld was fully isotropic was also simulated. The echo-dynamic results are presented
in Figure 144 showing that the amplitudes due to the use of the ADLs was consistently
higher than when the isotropic delay law was used but the echo-dynamic shape (compared
to what would be expected had the weld been isotropic carbon steel) was not achieved in the
anisotropic weld by either the adapted or isotropic laws.
The use of the flaw itself, which is specular to the beams, for generation of the ADLs is noted
as the most well-conditioned scenario and the increase in amplitudes due to the ADLs imply
that the sound field coherence is better maintained on incidence on the flaw. However, often
the nature of the flaw at a region of the weld is unknown and in such cases the ADLs would
need to be developed using the standard SDH reflector.

Ref: TP AA058J-FR
Version: 01.05
Figure 143 The simulated received amplitudes from Flaw 3 when using isotropic and
adapted delay laws; the array used 32 elements and was place in position P7.
Figure 144 The simulated echo-dynamic pattern from Flaw 3 when using the adapted and
isotropic laws (left), and in comparison to the expected shape had the weld been isotropic
carbon steel (right).
Figure 145 summarises the two different approaches which could be taken to generate the
ADLs through simulation: if the expected nature of the flaw is known and if a generic law
better optimised for a region of the weld is required.

Ref: TP AA058J-FR
Version: 01.05
Figure 145 Two possible approaches to generate ADLs through simulation: using the
expected flaw itself (left) or using a generic SDH in the correction region of the weld (right).
Figure 146 shows the ADL generated using the SDH from position P7 to those using the flaw
at three different positions (P1, P7 and P13) showing that the laws appear to be sensitive to
the microstructure. There is good confidence in the analytical theory (termed Kirchhoff) used
to evaluate the interaction of the sound field with both the specular crack-like Flaw 3 and the
SDH [20 and 21] when the medium is isotropic but thorough verification of their validity in a
simulated anisotropic medium is yet to be fully accomplished (see Section 7.4).
Figure 146 The ADLs generated using a SDH compared to those generated using Flaw 3
(from positions P1, P7 and P13) and with the isotropic law.

Ref: TP AA058J-FR
Version: 01.05
Scans over the flaw were then simulated again using the ADLs generated on the flaw, using
the SDH and using the calculated isotropic delay law. Further scan positions at 2mm
increments from P1 were added (P03, P05, P07, P09 and P011) which were all outside the
flaw (assuming isotropic straight beam incidence geometry). The resultant echo-dynamic is
presented in Figure 147.
Figure 147 The echo-dynamic from simulated scanning over Flaw 3 using ADLs generated
using a SDH, using Flaw 3 and using the isotropic delay law.
Firstly the actual peak echo strength using the isotropic delay law occurs outside of the
nominal incidence on the flaw, ie at position P03, and the peak echo is larger than those due
to the ADLs generated using either the flaw itself or a SDH. Note that the data between the
isotropic delay law and the flaw adapted delay laws (ie pink and yellow curves, respectively)
in the region from position P1 to P13 is the same as shown in Figure 144. The echo-dynamic
when using the ADL generated from a SDH (ie the blue curve) is closer to the expected
echo-dynamic from a flaw such as Flaw 3 and its profile is better than that from the other two
cases. All three cases did not position the maximum of the signal over the centre of the flaw
as would be required. Hence, in summary, the ADLs did not lead to an increase in the
maximum recorded echo strength but, in the case of the ADL generated using the SDH, the
echo-dynamic profile of the scan is better than when using the isotropic delay law. As
discussed in Section 6.1, the evidence supports the hypothesis that the ADL technique would
not be able to correct for positional errors.
Assuming that the incident beam is smaller than the flaw, using the 6dB drop method of
sizing the echo-dynamic of the data from the ADL using a SDH, the flaw size is measured as
12mm, a through-wall height of 8.5mm; the through-wall height of the simulated flaw was
8mm.
All the ADLs presented above were generated in the first section of the weld mapped using
EBSD (see Section 3). Figure 148 shows the difference in the ADLs generated using section
1 and section 2 of the weld microstructure.

Ref: TP AA058J-FR
Version: 01.05
Figure 148 ADL generated using a SDH at the position of Flaw 3 in weld sections 1 and 2.
The echo-dynamics due to the simulated scanning over Flaw 3 in the two weld cross
sections is presented in Figure 149.
Figure 149 Echo-dynamics over Flaw 3 when scanning is simulated in two weld sections.
Firstly, Figure 149 shows that the absolute amplitudes received from Flaw 3 in section 1 are
much higher than in section 2; the data from only section 2 is presented in Figure 150. The
performance of the ADL generated using a SDH in section 2 leads to lower received

Ref: TP AA058J-FR
Version: 01.05
amplitudes from the flaw, in comparison to the isotropic delay law. However, the amplitudes
due the use of both delay laws in section 2 are much lower than in section 1 implying
significant differences in attenuation due to the microstructures of sections 1 and 2. Note,
however, that the echo dynamic profiles over the actual flaw position are better in section 2.
Figure 150 Echo-dynamics over Flaw 3 when scanning is simulated in section 2.
Figure 151 shows the received amplitudes on each of the elements on the array when
generating the ADL on the SDH. Firstly, the data shows that, in comparison to section 1, the
echoes returning to the elements in section 2 is much reduced, implying severe relative
attenuation. Secondly, the data also shows that sound amplitudes returning to some of the
elements is higher than for others, implying that some elements contribute much less to
illuminating the region of the flaw.

Ref: TP AA058J-FR
Version: 01.05
Figure 151 Received amplitudes on each of the elements when generating the ADLs using a
SDH in weld sections 1 and 2.
Differences in the microstructure are further highlighted in Figure 152 through ray-tracing,
where the red and green rays emanating from the probe for the different elements represent
the theoretical shear and longitudinal wave paths, respectively. The path taken by the rays is
dependent on the properties of the medium, ie the microstructure, through which they travel.
Hence the results show (see the difference in behaviour of the rays from element 14 in
sections 1 and 2) that there are significant differences in the mapped sections. This
highlights the discrepancy between the experimental assessment of weld uniformity (see
Section 3.6) which concluded that the weld was relatively uniform and the results in Figure
152 which appears to suggest significant differences between the mapped welds. The issue
could be due to the mapping of the microstructure but there is also experimental evidence
(using the validation block, see Section 7.4) that the actual microstructure can be variable
depending on the angles at which the beams travels with respect to the dendritic grains.
Figure 152 Ray-tracings of shear (red) and longitudinal (green) waves emanating from
elements of the array in weld sections 1 and 2.

Ref: TP AA058J-FR
Version: 01.05
ADLs were also generated for inspection of Flaw 4 using SDHs placed at the position of the
flaw and one half of the TRL probe developed in the project (a 1.5MHz 32 x 2 array). In the
case of inspecting Flaw 4, as shown in Figure 153, the sound from the array has to traverse
the buttering layer, which was found to be difficult experimentally (see Section 7.2).
Additionally, the work done above on Flaw 3 made use of the 32-bit version 9 of CIVA
whereas the models for inspection of Flaw 4 made use of the 64-bit version 10 of CIVA.
Hence it was assumed that there were no changes in the mathematical calculations between
versions (eg in Kirchhoff interaction theory, refraction at boundaries etc) and it was found that
simulation times in version 10 were orders of magnitude quicker.
Figure 153 Generating ADLs for inspection of Flaw 4 in weld sections 1 and 2.
The Kirchhoff theory was used to generate the ADLs using a SDH but another theory termed
the Geometric Theory of Diffraction (GTD) was used for evaluating the interaction of the
sound with Flaw 4. Kirchhoff theory has been validated for use near specular incidence [20]
but it is not able to deal with scenarios where the interaction will give rise to diffracted
echoes, for which GTD is better suited. GTD has been validated for use in many of the
models developed by British Energy [20] but the limits to its validity as utilised in CIVA, in
particular when used in the anisotropic medium of the weld, has not been fully established.
Figure 154 shows the delay laws generated in section 1 (ADL 1) and section 2 (ADL 2) in
comparison to the isotropic delay law calculated by CIVA and ArrayGen. There are in total 64
elements in the array (represented on the x-axis) and the magnitude of the delays for the
elements (represented on the y-axis) implies that the isotropic velocities assumed for the
calculated laws - Iso (CIVA) and Iso (AG) – closely matches the velocities in the anisotropic
weld, as expressed through the stiffness coefficients. The profile of ADL 1 is similar to those
of the isotropic delay laws but ADL 2 suggests severe variability between elements. The
profile of the ADL in section 2 is again indicative of (1) the complexity of the microstructure
and (2) its difference when compared to the microstructure of section 1.

Ref: TP AA058J-FR
Version: 01.05
Figure 154 ADLs generated using a SDH in weld sections 1 and 2 for inspection of Flaw 4.
The received signal amplitudes on each element of the array are presented in Figure 155 for
each of the weld sections. The received elemental signal amplitudes are higher when
generating the ADL in section 2 but the distribution of the amplitudes again indicates the
degree of distortion being induced by the microstructure of the second weld cross section in
comparison to the first weld cross section.
The scanning for Flaw 4 was then simulated, as in the case of Flaw 3, and the results are
presented in Figure 156 for section 1; note that the theory used for the interaction of the
sound with the flaws was GTD. The results appear to be consistent with the experimental
data discussed in Section 7.2, in terms of the loss in amplitudes, differences in plotting,
sizing and the effects due to distorted sound fields. In the anisotropic weld the isotropic delay
law plots the echoes from the flaw inside the weld whereas the ADL plots the flaw outside the
weld and both suggest spatial distortion due to the flaw tips. The results of inspection through
section 2 are similar, as presented in Figure 157. In both cases (ie inspection through
sections 1 and 2) the method by which CIVA presents the scan data makes it difficult to
identify diffraction signals and hence size the flaw; note that in the case of the isotropic delay
law being used to inspect in an isotropic weld, the diffracted signals can be clearly resolved
allowing for the flaw to be sized.
In summary, in section 1, the ADL profile (shape and amplitude) is similar to the two isotropic
delay laws generated by CIVA and ArrayGen; however, only 10 elements contribute
significantly towards illuminating the flaw using the ADL. The improvement in signal
amplitude when using the ADL is estimated to be less than 2dB; the simulated position and
size of the flaw using the ADL is also larger than when the isotropic delay law was used. The
ADL generated in section 2 suggests significant local microstructural variation and the
received amplitudes from each element also suggest significant differences in attenuation.

Ref: TP AA058J-FR
Version: 01.05
Figure 155 Received signal amplitudes on each element of the array when generating the
ADL using a SDH in each of the two weld sections.
Figure 156 Simulated scanning results for inspection of Flaw 4 using the GTD theory of
interaction; isotropic weld, isotropic delay law (top left), anisotropic weld, isotropic delay lay
(top right) and anisotropic weld, ADL (bottom).

Ref: TP AA058J-FR
Version: 01.05
Figure 157 Simulated scanning results for inspection of Flaw 4 using the GTD theory of
interaction; isotropic weld, isotropic delay law (top left), anisotropic weld, isotropic delay lay
(top right) and anisotropic weld, ADL (bottom).
As the sound travels through the weld it traverses several ‘grains’ or major orientation
regions. As discussed in Section 3, the orientation unification processing leads to the
selection of several major orientations with the misorientation parameter selected such that
the whole weld can be described and mapped using a limited number of regions assigned
one of the major orientations. The size of the regions, and hence the misorientation
parameter and number of major orientations, is governed by placing the inspection within the
Rayleigh scattering regime (minimum scattering criterion). A further route to simplify the weld
description was explored to follow the orientation unification method in order to limit distinct
regions and hence reduce simulation times. The aim was to consider the behaviour of the
sound beam as it passes from a region of one major orientation into another one.
The minimum scatter criterion for the weld specimen and inspection frequencies of the
project states that the minimum size of the ‘grains’ or regions will be 300µm but they could be
larger. The propagation of the sound by the model is subject both to the orientation of these
regions (termed domains for the study) and their size. Figure 158 describes the two tests
performed as part of the study, one test to investigate orientation changes and the other to
investigate changes in the size of the regions or domains. Four major orientations were
chosen: red, lime green, yellow and blue. The sound was generated by a single element
25.4mm in diameter. The path of the longitudinal wave is represented by the green line in
Figure 158, deflecting from domain to domain due to changes in property. The model is used
to calculate the beam cross section at as the beam emerges from its travel into the last
domain.

Ref: TP AA058J-FR
Version: 01.05
Figure 158 Tests to study the effect of changes in the orientation and sizes of distinctly
different regions (domains).
The two primary orientations found in the buttering are red and yellow. Figure 159 shows the
effect on the propagating sound when travelling between these two different orientations;
note that the isotropic domain case is when Domains 1 and 2 in Figure 160 Test 1 are
isotropic.
Figure 159 Study of the influence on the propagating sound beam due to changes in the two
major orientations found in the buttering.
The key parameters are the cross beam width (ie changes in size) and changes in position
with respect to the datum, which is the origin of the x-y plane. Red to red leads to the
greatest increase in beam size and yellow to yellow leads to the greatest shift in position.

Ref: TP AA058J-FR
Version: 01.05
The beam travelling from red to yellow and from yellow to red lead to the same level of
deflection and change in beam size. Compared to the reference isotropic scenario the
change in amplitude is all cases is severe leading to a loss of about 18dB but remains similar
in all cases involving the orientations.
Within the weld body itself orientations represented by lime green and blue are also largely
present in addition to red and yellow. Figure 160 shows the orientation study (Test 1) in the
body of the weld section 1. The losses in amplitude in comparison to the reference isotropic
cases are similar to the buttering case, about 18dB.
Figure 160 Study of the influence on the propagating sound beam due to changes in the two
major orientations found in the weld.
Changes in amplitude compared to the isotropic cases for the different possible scenarios
are presented in Figure 161. Similarly the changes to the beam on the x-y plane (termed
misorientation) with respect to the isotropic case are presented in Figure 164.

Ref: TP AA058J-FR
Version: 01.05
Figure 161 Summary of changes in amplitude in the different scenarios (Test 1).
Figure 162 Summary of changes in the beam direction with respect to the isotropic case in
the different scenarios (Test 1).
The Test 2 scenario using the red and green orientations is presented in Figure 163 for
differing thicknesses of the sandwiched Domain 2 (see Figure 158). The results show that
there is no significant change in the deflection of the beam or its size when the sandwiched
region is less than 1mm, in comparison to the case where the sandwiched region is 1mm.

Ref: TP AA058J-FR
Version: 01.05
Figure 163 Study of the influence on the propagating sound beam due to changes in the
thickness of a green region between sandwiched between red regions (Test 2).
A similar result was found in the case of a red region sandwiched between two yellow
regions, as shown in Figure 164, where there is no significant change to the propagating
beam when the region is less than 1mm, in comparison to the case when the sandwiched
region is 1mm.
Figure 164 Study of the influence on the propagating sound beam due to changes in the
thickness of a red region between sandwiched between yellow regions (Test 2).

Ref: TP AA058J-FR
Version: 01.05
The results of the limited study presented here shows that changes in amplitude are not
significantly affected by the changes in domain; however note that CIVA does not take into
account scattering losses (see Section 7.4). Since scattering losses are not calculated in
CIVA, it could be possible to increase the minimum region size to 1mm as regions smaller
than this do not seem to impact the propagation of the sound (Test 2). Hence smaller regions
will be merged into the closest, largest and most similar region.
Parametric studies such as the examples presented here should lead to a greater
understanding of the sound propagation in the inhomogeneous anisotropic medium, possibly
leading to further processing methods to aid in reducing the modelling times while capturing
the propagation characteristics sufficiently.
7.3.2 Experimental
The ADLs developed, as described in Section 7.3.1, were then used to scan the specimen
and compare with scans using the conventionally calculated isotropic delay laws. Figure 165
shows the experimental setup with the probe described in the simulations sitting on the
DISSIMILAR specimen and connecting (dashed red circle) to the British Energy X-Y
scanning frame to implement automated scanning using MIPS. The collected data was then
processed in GUIDE for presentation in this document.
Figure 165 The 2MHz linear array probe being connected to the British Energy scanning
frame to undertake experimental validation of the ADLs generated for Flaw 3.
Line scans were performed to recreate the simulated scans over Flaw 3 described in Section
7.3.1 using both the ADL generated using a SDH and isotropic delay laws. Figure 166 shows
the experimental echo-dynamic over Flaw 3 due to the ADL and the calculated isotropic
delay law. The peak value of the echo-dynamic using the ADL was about 1dB smaller than
that using the isotropic delay law. However, the 6dB drop size provided by the ADL is closer
to the size of the actual flaw in comparison to the size measured on the isotropic echo-
dynamic.

Ref: TP AA058J-FR
Version: 01.05
Figure 166 The experimental echo-dynamics over Flaw 3 using the adapted and isotropic
delay laws.
There are several issues to address between the simulated weld and the actual weld:
The actual stiffness coefficients of the particular alloy used in the weld was not evaluated
using single crystals (see Section 7.4)
The simulated weld was smaller due to loss in thickness due to cutting of coupons for
EBSD scanning
Lengthwise microstructure variations between the section used to generate the ADL and
the actual weld region where the flaw is present
Experimental evidence (see Section 3.6) showed that microstructural variation along the
length of the weld did not significantly affect amplitude or positioning of echoes generated by
Flaw 1. However, simulated evidence (see Section 7.3.1) suggests that the ADLs being
developed appear highly sensitive to changes in microstructure between EBSD mapped weld
sections, and between domains of differing orientations and sizes. In effect, the results
presented in Figure 166 suggest that no significant improvements in received amplitude
could be achieved through the use of ADLs, contrary to the simulated evidence.
The ADLs generated in Section 7.3.1 at the different positions were then used to implement
a position dependent inspection, ie at a given position the corresponding ADL was fired. To
execute this inspection concept, the extended flaw file capability of GUIDE was utilised such
that the actual position dependent inspection was implemented virtually during processing in
GUIDE. Figure 167 shows the positions of the simulated inspection which was recreated
experimentally. Assuming the sound beams travel in a straight line (which is inherent within
the GUIDE processing system), the echo received from the ADL at its corresponding
inspection position is stitched together next to the other positions within GUIDE

Ref: TP AA058J-FR
Version: 01.05
Figure 167 The positions over Flaw 3 where ADLs were generated and which are then used
to implement a position dependent inspection virtually within GUIDE.
Data is collected over all the positions in Figure 167 using all the ADLs and the results of the
scans are shown in Figures 168 and 168.
Figure 168 Results of scans over Flaw 3 using ADLs generated for positions P13, P11, P9,
P7, P5 and P3.

Ref: TP AA058J-FR
Version: 01.05
Figure 169 Results of scans over Flaw 3 using ADLs generated for positions P1, P03, P05,
P07, P09 and P11.
The scans using each of the ADLs have variable performance, some better than others in
detecting the flaw with good S/N and received echo amplitudes; ADLs generated at positions
P13 and P11 were particularly poor in detecting the flaw.
Figure 170 describes the process of executing a position dependent scan using the data
shown in Figures 168 and 169 within GUIDE.
Figure 170 Executing a position dependent scan virtually using a group of MIPS scan data,
selecting the relevant data and stitching the scan image back together.

Ref: TP AA058J-FR
Version: 01.05
Figure 171 shows the data from scanning using the isotropic delay law and Figure 172
shows the results from the position dependent scan using all the ADLs generated using Flaw
3 (see Section 7.3.1).
Figure 171 The scan over Flaw 3 using the calculated isotropic delay law.
Figure 172 The scan over Flaw 3 using the generated ADLs using Flaw 3.
The experimental validation results again show that the use of the ADL does not lead to a
significant increase in sensitivity to the flaw as suggested by simulation results (see Section
7.3.1). Hence in summary, the evidence generated in the DISSIMILAR project suggests that
the concept of generating ADLs using simulations is viable, in that it gives similar
performance to conventional phased array techniques, but the use of ADLs does not appear
to increase the sensitivity of the inspection to flaws which lie specular to the incident beam.
The time and budget only allowed for a limited study and further investigations on other flaws
in different orientations and different parts of the weld should be undertaken in future efforts.

Ref: TP AA058J-FR
Version: 01.05
A study was conducted to generate ADLs experimentally, ie without using models and
simulations. The probe developed in a previous project and described in Appendix A was
used along with a British Energy reference block in immersion. Figure 173 shows the
reference block containing the SDH target which was used as the target to generate an ADL
and the probe which is a TRL type of 2.5MHz with 16 x 4 elements on each lobe with a pitch
of 2mm x 3mm, respectively, including a wedge angle of 13˚ built into the housing.
Figure 173 The reference block and TRL probe used for experimental generation of ADLs.
The method was to fire on an element in the transmit lobe of the probe and listen to the
echoes on a different element in the receive lobe of the probe. Figure 174 shows the A-scans
recorded on a selection of transmit-receive pairs; the probe was placed on the stainless steel
side targeting the SDH – ie the sound did not penetrate the weld.
Figure 174 A-scans recorded for a selection of transmit-receive pairs.
Identifying the echo from the SDH proved to be difficult even in the relatively clean material
of the reference block. The expected position of the echo due to the SDH was calculated but
the S/N quality of the signals was judged to be poor. The quality of the signals, in particular
at the relatively high frequency of 2.5MHz, when the probe was used to detect slot EF (see
Figure 173) through the weld was severely degraded. Hence the ADLs could not be
generated; however further investigation of experimental methods similar to the limited study
presented here warrant further investigation (see Section 10.2).

Ref: TP AA058J-FR
Version: 01.05
7.4 Model validation
The CIVA model was used as the primary platform for the simulation of inspections and
generation of the ADLs (see Sections 6.2 and 7.3.1). Its use in isotropic media for use as
conceived in the DISSIMILAR inspections has been validated by TWI [20, 21] and
elsewhere. The key aspect of the model, its treatment of sound propagation in
inhomogeneous anisotropic media, has not yet been fully validated. The intention was to
validate the model using a specially grown single crystal made of the alloy used in the
DISSIMILAR specimen weld. However, the attempt at growing the single crystal was not
successful and hence an alternative method using the weld itself was explored.
In a single crystal of an alloy the unit cubic crystal is aligned in a specific direction. Hence the
elastic properties of the macro single crystal specimen are a function of the unit cubic crystal.
Importantly for validation, the single crystal specimen can then be introduced into the model
as a homogenous medium which takes the cubic stiffness constants of the unit cubic crystal.
Then the simulated propagation of sound (measuring amplitudes and spatial field
characteristics) can be verified against experiments identically recreating the simulations.
Methods of growing single crystals have been developed recently for fabricating single
crystal turbine blades for their superior mechanical properties. The Bridgeman growth
process (see Figure 175) involves generating a wax model which is used to create an
investment shell to create a mould. The molten metal is then introduced into the mould at
temperatures of around 1500˚. The crystal is then grown slowly as the liquid-to-metal
interface is slowly moved. The grains grow along the <001> crystallographic direction since
the unit cube is FCC.
Figure 175 The Bridgeman single growth process for a single crystal turbine blade (image
courtesy of Rzeszow University of Technology, Poland).

Ref: TP AA058J-FR
Version: 01.05
Figure 176 shows the cylindrical rod of diameter 20mm and length of 70mm which was
grown using TIG consumable rods of the Inconel alloy used as the DISSIMILAR specimen
weld filler. Analysis in SEM showed that that the growth was now successful and differential
boundaries had been introduced (see Figure 177). It was deemed that the specimen would
not be used for the validation or for evaluating elastic stiffness constants (see Section 3.7)
Figure 176 The Inconel alloy single crystal cylinder (20mm diameter, 70mm length) grown
for the validation of the model in the DISSIMILAR project.
Figure 177 The SEM scan of the single crystal rod on the root side (left) and possible
boundaries within the specimen (right).

Ref: TP AA058J-FR
Version: 01.05
An alternative method was then explored using a coupon from the reference weld block
extracted from the DISSIMILAR specimen. Figure 178 shows the extraction of the coupon
into which five 3mm SDHs were introduced. The dimensions and target SDHs were similar to
a standard carbon steel reference block shown in Figure 179. Both the weld coupon
(hereafter termed the validation block) and the carbon block were then scanned using a
12.5mm diameter 3.5MHz immersion probe in a scanning tank; the concept is described in
Figure 180.
Figure 178 Extraction of a coupon from within the weld towards validation of the model.
Figure 179 The reference carbon steel calibration block containing target SDHs and
dimensions similar to the coupon extracted from the weld (see Figure 175).

Ref: TP AA058J-FR
Version: 01.05
Figure 180 The validation block and the carbon block are raster scanned in an immersion
tank using identical probe (green block above) and standoff distances; the system gain value
is selected to be able to set the back wall signal to near 80% FSH.
The scan of the carbon block is shown in Figure 181 and the scan of the validation block is
shown in Figure 182.
Figure 181 A scan of the carbon steel block showing three SDHs with equal echo
amplitudes, equidistant from each other at the same depth from the top surface, consistent
with the actual physical positions shown in Figure 176. The A-scan data is along the dotted
blue line through one of the target SDHs.

Ref: TP AA058J-FR
Version: 01.05
Figure 182 A scan of the validation block showing the effects of sound field distortion with
unequal amplitudes from the targets, errors in positioning, increased attenuation (note the
increase in gain required to put the back wall signal to 80% FSH) and much reduced S/N
performance. The dotted red lines indicate positions of the target SDHs and the A-scan data
is along the leftmost line.
The aim of the validation effort is to check the results from the simulation predictions against
the data presented in Figures 181 and 182. The scan performed in the tank was then
simulated using CIVA and the carbon steel block and the results are shown in Figure 183.
The result of the simulation is consistent with the experimental data given in Figure 183:
equal amplitudes from the target SDHs and the positioning of the echoes are consistent.
The corresponding simulation using the validation block is presented in Figure 184 where the
weld microstructure is introduced as the weld medium from the two EBSD mapped weld
sections. The echoes returned from the target SDHs are no longer coherent and appear to
be subject to the type off distortion presented in Figure 182. Hence qualitatively, the model
appears to be simulating the effects observed in reality. To undertake the validation of the
model, the returned echoes in the simulation must be compared to the experimental echoes
generated in the actual validation block; however, the actual microstructure of the
experimental validation block is different (by an unknown degree) to the simulated
microstructure. This gives rise to an error or inaccuracy between the simulated and
experimental cases, and furthermore this error is difficult to quantify.
Figure 185 shows, for the two weld sections, the simulated signals returned from a SDH
target in the carbon steel block (same signal was received from all three) and the signals
from the three SDHs in the validation block along the dotted red lines shown in Figure 182
are plotted at the bottom. The relative difference in signal strength between the SDHs in the
carbon steel and the validation block is (42.5-10)dB + 13dB = 45.5dB.

Ref: TP AA058J-FR
Version: 01.05
Figure 183 Simulation result of the immersion inspection of the carbon steel block.
Figure 184 Simulation result of the immersion inspection of the validation block. The dotted
red lines show the position of the three middle SDHs and the images below show the
microstructural maps of the two sections input into the model to simulate the validation block.

Ref: TP AA058J-FR
Version: 01.05
Figure 185 Simulated signals from the SDHs in the carbon steel (top) and the validation
block (bottom) for the two different weld sections.
To complete the validation, the simulated signals shown in Figure 185 must be compared to
the A-scan data shown in Figure 182. However, note the absence of the grain scattered
echoes evident in Figure 182 but not in the simulated signals of Figure 185. This is because
the CIVA model is known to not take into account attenuation due to scattering (hence the
absence of noise and only signals due to introduced targets). There will therefore be a large
discrepancy between the predicted amplitude levels and the actual attenuation as scattering
is the single largest attenuation mechanism in coarse grained materials [2]. Hence validation
of the CIVA model using amplitudes of received signals is not relevant.
In terms of the times-of-flight to targets predicted by the model there is good agreement
between the results of the simulations and the experimental data using the validation block;
the discrepancies measured in time which are less than errors in positional differences
between reality and the simulations (<1mm). The result also explains the remarkable
similarity in amplitudes and profile shapes between the calculated delay laws and the
simulation generated ADLs, as shown in Figures 154 and 166. The evidence suggests that
the use of the CIVA model is suitable for the generation of the ADLs which is dependent only
on accurately simulating the times-of-flight. However it would not be suitable for use in
predicting signal amplitudes and hence studies for setting sensitivities during inspections in
such inhomogeneous coarse grained materials.
At the commencement of the project the CIVA model was chosen as the primary simulation
platform because of the ease with which inspection concepts can be realised within its
dedicated framework. However finite element models were also investigated and a
discussion of their capabilities is presented in Section 10.3 and demonstration of their ability
to generate ADLs as wells as being able to simulate grain scattering effects.

Ref: TP AA058J-FR
Version: 01.05
8 Conclusions
1. The performance of the phased array techniques in detection, positioning and sizing was
not appreciably better than conventional ultrasonic techniques.
2. The signal-to-noise performance of the transmit-receive longitudinal (TRL) technique was
substantially better than that of the other techniques.
3. The ability to skew the beam allows for better determination of the three dimensional (3D)
orientation of a flaw; the use of a two dimensional (2D) array allows a range of skewed
beams to be generated at increments smaller than 1˚ enabling the measurement of
defect tilt and/or skew (ie 3D orientation). The same precision would require multiple
conventional probes generating each skewed beam individually, leading to greater costs.
4. The primary justifications for using phased array techniques include:
Versatility; one probe can generate a range of beams and hence perform the function
of numerous conventional probes that are each able to generate only a single beam.
Inspection time; phased array techniques reduce the total time for inspection and
hence reduce the overall costs through increasing operating revenue.
Dosage; reducing the inspection times also leads to a reduction in the time personnel
will be exposed to radiation, presenting a strong safety case for the use of phased
array technology in radioactive environments.
Debris; the likelihood of introducing debris into the containment area is reduced as
phased array systems (through their versatility) can be designed to be less
complicated to deploy mechanically.
Digital data archiving; data can be digitally stored for future analysis and reference.
5. The impact of the microstructure, in particular its variation along the welding direction, will
dictate whether the modelling route to generating ADLs can be implemented cost
effectively. In reality, it is unlikely that the entire weld volume can be quantified accurately
and cost effectively; additionally, the weld used for quantification would not be the same
as that inspected (assuming the quantification process is destructive). Hence there will be
a degree of variation between the weld being modelled and the actual one being
inspected. If the degree of difference is too large then the ADLs generated will not be
applicable for inspection.
6. The EBSD technique was successfully used to map a cross section of the dissimilar
weld, identifying different areas (such as buttering and cladding) and, through the use of
a processing method developed in the project termed orientation unification, allowed the
weld to be quantified in the ultrasonic model.
7. The limited work done in the project shows no appreciable benefit, in terms of detection
and sizing of flaws, in using the ADL technique. The increases in received signal
amplitude predicted theoretically could not be verified experimentally; reasons for this
may include not being able to sufficiently capture the actual microstructural condition (see
6 above) and/or inaccuracies in the ultrasonic model (which has not been fully validated
for this application).
8. The evidence based on the validation block shows good qualitative correlation between
prediction and experimental data in the sound distortion in the coarse grained weld
material. However, the CIVA model used for generating the ADLs was not fully validated
as the single crystal specimen intended for this purpose was not successfully grown.

Ref: TP AA058J-FR
Version: 01.05
9 Recommendations
1. To achieve effective inspection of coarse grained, inhomogeneous and anisotropic

material, the signal-to-noise performance of the ultrasonic system must be optimised.
Hence techniques making use of transmit-receive longitudinal probes are recommended.
2. Two dimensional array probes are recommended for determining the three dimensional
orientations of flaws.
3. The use of validated modelling tools to explore the capability of the phased array
techniques being developed is highly recommended, as models offer a cost effective
route for designing, checking and eliminating potentially costly shortcomings in the
inspection effort.
4. The ADL technique developed in the project has not been sufficiently developed for
general use; any future implementations must be subject to qualification evidence. The
method shows potential to improve detection capabilities theoretically and may provide
significant benefits in future.
5. The models being used for generation of ADLs but also to simulate inspection in the
inhomogeneous anisotropic weld media require stringent validation and care should be
exercised when using models whose limitations in accurately simulating physical
phenomenon have not yet been established.

Ref: TP AA058J-FR
Version: 01.05
10 Future directions
The work done in the DISSIMILAR project was aimed at the evaluation of the current state-
of-the-art with regard to the tools available for the inspection of the particular case of coarse-
grained inhomogeneous and anisotropic austenitic welds. The following three aspects were
identified as areas under development which may provide a way forward in the near future
and improve present ultrasonic inspection capabilities for this special class of fusion joints.
10.1 Array probe capabilities
During the course of the project several aspects of probe manufacture were identified as
critical to fabricating probes with the characteristics identified in Section 4.2. The goal is to
increase the bandwidth of the signal to above 80% of the centre frequency, which requires
mastering manufacturing processes such as lapping, electroding methods, matching layer
thicknesses and selecting material with the right characteristics for damping. These aspects
are commercially sensitive and will not be discussed further in this document.
Another important aspect of probe technology is enhancing the capabilities offered by

different array configurations. Annular array configurations offer the possibility to focus the
sound energy to smaller volume, leading to smaller beam sizes. This offers the possibility to
increase the S/N performance and also improve the sizing capability of techniques. Figure
186 shows three annular arrays (one pure annular with rings only and two with segmented
rings) whose sound field characteristics were studied through simulations.
Figure 186 Annular arrays investigated for sound field characteristics.
All the arrays were of 2MHz frequency and were used to focus at 100mm depth from the
probe face (the beam widths were measured at a distance of 94mm from probe face). The
diameter of all the arrays was 30mm with a near field range of 437mm implying the focusing
was to a factor of 0.23. The beams were steered to 0, 0.6, 1.7, 2.9, 5.7 and 8.5˚ and the 6dB
beam width was then measured, as shown in Figure 187.

Ref: TP AA058J-FR
Version: 01.05
Figure 187 Simulated sound fields when the beam is steered up to 8.5˚ using the three
different annular array configurations.
The results show that with a 128 element segmented annular array very tight beam spots
can be maintained at high steering angles whereas the sound field integrity fails at lower
steering angles when using both the low element count segmented annular array and the
pure annular array. This array configuration should be investigated in future for application to
the inspection of dissimilar / austenitic welds.
Another potential array configuration (introduced in Section 4.6) which shows potential for
greater manipulation of the 3D sound field is based on aperiodic fractal designs [22]. Fractal
designs offer the possibility to achieve similar performances to ‘conventional’ configurations
but with a much lower element count. Such sparse arrays, based on fractal shapes, have
been shown to offer remarkable focusing and steering performance. Figure 188 shows two
sparse array designs with a total number of elements less than 128 but over a large area and
Figure 189 shows inspection data from a prototype spiral array design probe.
Further concepts which should be investigated in the future include:
1. Hard wired 1.5D array configuration: When the beams are required to be steered in only
one plane, then a fully 2D array configuration could be replaced by a 1.5D concept where
the symmetry between the left and right halves can be electronically expressed by
physically connecting them to the same wires, as described in Figure 190. This exploits
the fact that the delay laws on either side of the line of symmetry are identical and is
ideally suited when the secondary axis is used for focusing.
2. Two way transmission using TRLs: In the TRL technique explored in the project the
sound was transmitted from one half of the TRL probe and received on the other. The
sound travels through a scattering medium but reciprocity may not be maintained. Hence
if the travel path of the sound is reversed, then the grain structure may affect the
propagation differently. Hence by transmitting in both directions and using signal
processing techniques inspection capabilities could be enhanced (see Figure 191).

Ref: TP AA058J-FR
Version: 01.05
Figure 188 Two examples of sparse array configurations based on fractal designs.
Figure 189 Inspection results from a prototype spiral array design probe; the TFM method is
described in Section 10.2.

Ref: TP AA058J-FR
Version: 01.05
Figure 190 The hard wired 1.5D array configuration which is only feasible when no steering
along the secondary plane is required.
Figure 191 Transmitting both ways using a TRL configuration as the effect of the grain
structure on the propagating sound wave may be different in both directions.
10.2 Inspection techniques based on full matrix capture of data
Post-processing techniques making use of the full matrix capture (FMC) of data are an
avenue for exploration in future efforts. In FMC, each element on the array is fired and the
signal received on all elements of the array is recorded. The result is a data matrix that
contains all possible transmit-receive combinations on the array. This data matrix can then
be manipulated, applying virtual delay laws, such that techniques can be implemented
virtually. Techniques based on FMC have been shown to have advantages over conventional
phased array techniques for inspection of isotropic carbon steel material [23].
The use of techniques based on FMC for austenitic materials has not been widely explored
at the compilation of this document. In the DISSIMILAR project a brief study was conducted
using the specimen presented in Appendix A and shown again in Figure 192. A linear array
probe of 2.25MHz was used with an 18.5˚ Rexolite wedge to generate longitudinal waves
within the specimen. The FMC data was captured and a post-processing technique termed
the Total Focusing Method (TFM) [23] was used to image the data. The TFM algorithm sums
together at each imaging point, ie the point within the component, all the received signal
amplitude for all transmit-receive pairs within the data matrix. The underlying assumptions of
the post-processing technique are that the propagation of sound from the transmitter to the
receiver is along straight lines (subject only to refraction at the component / wedge boundary)
and that the velocity is constant.

Ref: TP AA058J-FR
Version: 01.05
Figure 192 Specimen used for investigating the use of FMC techniques (see Appendix A).
Inspection was performed with the probe placed on either the carbon (ferritic) steel side or
the stainless steel side and looking at the two slots H and L shown in Figure 192. Figure 193
shows the results from imaging the reference slot which was used to calibrate the technique;
note that the imaging amplitude is automatically normalised to the highest amplitude in the
imaging region by the processing algorithm.
Figure 193 Image of the reference slot in the specimen showing the back wall, block corner
and tips of the slot (compare to images A10 – A12 in Appendix A).

Ref: TP AA058J-FR
Version: 01.05
Figure 194 shows imaging of slots H and L from the stainless steel side with the probe
placed at four different positions which lay close to each other. The data shows that the level
of distortion is similar to that observed in the validation block (see Figure 182) and the
performance is similar to the other ultrasonic techniques (see Section 7).
Figure 194 Images of slots H and L in the specimen with the probe placed on the stainless
steel side of the weld (compare to images A13 – A19 in Appendix A).
The grain scatter gives rise to a significant level of back scattered echoes which manifest
themselves in the image after the TFM processing. Additionally the echo amplitudes received
from the slot tips fluctuate for the different positions which are not significantly spaced apart.
The probe was then placed on the carbon steel side and the slots were again imaged, this
time the sound having to traverse the buttering layer. Figure 195 shows images of the bottom
slot from the carbon steel side at two differing but nearby positions and Figure 196 shows an
image of the top slot from the carbon steel side. In both figures the manifestations of sound
field distortion as observed in other ultrasonic techniques is evident.
This study has shown that post-processing techniques (such as TFM) which make use of
FMC data will suffer similarly to other ultrasonic techniques unless the microstructural
conditions are accounted for in the processing algorithms. This will inevitably require the
weld microstructure to be described in some fashion to the algorithms so that they can
process the data to take into the effects due to the inhomogeneity and anisotropy.

Ref: TP AA058J-FR
Version: 01.05
The evidence to date shows that where ‘active’ phased array is able to penetrate a given
volume of weld, then post-processing methods using FMC data will also work. However,
future investigations should consider in depth the relationship between sound frequency and
attenuation for implementation of FMC because in FMC the sound wave from a single
element is required to penetrate the weld, reach the target and be received back on all
elements of the array. Elements of typical arrays are small and the sound emanating from
them is highly divergent, which may require special considerations for probes intended to be
used for FMC of data, leading to modifications in the probe specification methods described
in Figure 86.
Figure 195 Images of slot L (near the root) in the specimen with the probe placed on the
carbon steel side of the weld (compare to image A23 in Appendix A).
Figure 196 Image of slot H (near the top surface) in the specimen with the probe placed on
the carbon steel side of the weld (compare to images A20 – A22 in Appendix A).

Ref: TP AA058J-FR
Version: 01.05
10.3 Finite element modelling packages
Two finite element (FE) packages were investigated during the course of the DISSIMILAR
project: PZ Flex with a time domain solver and the general purpose ABAQUS using a variety
of solvers. At the commencement of the project it was decided that the FE packages did not
have the tool boxes available to adequately describe a complex inspection scenario (to
include probes, wedges, components, techniques etc) which remains the case at the
conclusion of the project. However, the inability to deal with grain scattering by the CIVA
platform was identified as a short coming and the simulation times with CIVA (both the 32-bit
version 9 and the 64-bit version 10) were deemed to be limiting factors to introducing the
novel ideas generated into wider industrial use. Hence FE models were investigated primarily
to establish whether it was possible to describe the weld microstructure into them and
whether it was possible to generate ADLs.
PZ Flex was used to model a simple inspection scenario over the DISSIMILAR weld using a
linear 1.5MHz array composed of 128 elements, as shown in Figure 197 [24]. A particular
advantage of using PZ Flex was the introduction of the weld microstructure map which was
done quickly and cost effectively by scanning in a colour map and assigning the orientations
to each colour. In CIVA the orientations assigned to each region was input manually taking a
considerable amount of time (estimated to be 3 working days).
Figure 197 The inspection case modelled in PZ Flex using the weld microstructure mapped
using EBSD, introduced using colour coded maps
Introduced into the middle of the weld was a 3mm SDH target. Then FMC data collection
was simulated (taking 90mins to complete), followed by a simulated implementation of the
TFM post-processing method (taking 60secs per image). The results after the TFM are
presented in Figure 198 which shows both the distorted echo from the target SDH as well as
the scatter due to grains (which could not be evaluated in CIVA). The FMC data was then
processed using the concepts behind generating ADLs and the result is presented in Figure
199 which shows that the amplitude and integrity of the echo from the target SDH is imaged
better, implying the integrity of the sound field is better maintained. Note, however, that the
background grain scatter noise level was not suppressed by the modified processing.

Ref: TP AA058J-FR
Version: 01.05
Figure 198 Image of the weld containing the target SDH using a virtual application of TFM.
Figure 199 Image of the weld containing the target after using time reversal concepts to
adapt the processing.

Ref: TP AA058J-FR
Version: 01.05
ABAQUS was used to model the validation block presented in Section 7.4 and was used to
generate an ADL to investigate the possibility to use the model as an ‘anisotropic delay law
calculator’. ABAQUS is a general purpose FE code and is hence open to configuration. A key
advantage presented by the package is the ability to code in the EBSD map at its full 40µm
resolution. This will eliminate the orientation unification processing step to generate a weld
microstructure map where the boundaries are set through assumptions. The raw EBSD map
(see Figure 27) is input (as a text file) directly into ABAQUS and the raster grid used by the
EBSD scanner will become the FE grid, containing millions of elements and nodes.
Figure 200 shows the concept where a linear array probe containing 32 elements is used to
scan the validation block, which in the first instance is introduced into the model using the
boundaries generated after the orientation unification step.
Figure 200 The scenario introduced to the ABAQUS model; a 32 element 4MHz linear array
is used for the inspection, focusing the sound to the position of the target SDH.
Figure 201 shows the propagation of the wave front from a single element when the
validation block is considered to be isotropic (image is at a time just before impact on the
SDH) and Figure 202 shows how the integrity of the wavefront is broken when it travels
through the anisotropic weld structure.
Note that in both Figures 201 and 202 the blue region represents the validation block, ie the
medium. Note also that, like PZ Flex and unlike CIVA, ABAQUS is able to evaluate back
scattered grain echoes. At the time of writing, work is ongoing to use ABAQUS to implement
a validation of the ADL concept, the model and the stiffness coefficients.

Ref: TP AA058J-FR
Version: 01.05
Figure 201 Wave front from a single element of the array just before impact on the SDH; the
validation block was assigned isotropic carbon steel values.
Figure 202 Wave front from a single element of the array just before impact on the SDH; the
validation block was assigned the orientation information as mapped by EBSD.

Ref: TP AA058J-FR
Version: 01.05
11 References
1. IIW Handbook on the Ultrasonic Examination of Austenitic and Dissimilar Welds, DVS Media
GmbH, Düsseldorf 2008.
2. Krautkrämer, J. and Krautkrämer, H., ‘Ultrasonic Testing of Materials’. 4th Ed., Springer-Verlag
(1990).
3. Baikie B. L., Wagg A. R., Whittle M. J. and Yapp D., ‘Ultrasonic inspection of austenitic welds’.
Journal of the British Nuclear Energy Society, Vol 15 No 3, 1976.
4. Juva A. and Haarvisto M., ‘On the effects of microstructure on the attenuation of ultrasonic waves
in austenitic stainless steels’. The British Journal of Nondestructive Testing, Vol 19 No 6, 1977.
5. Randle V. and Engler O., ‘Introduction to Texture Analysis’. CRC Press, 2000.
6. Papadakis, E. P., ‘Revised grain-scattering formulas and tables’. J. Acoust Soc Am., Vol 37, pp
703-710, 1965.
7. Juva A. and Lenkkeri J., ‘The effect of anisotropy on the propagation of ultrasonic waves in
austenitic stainless steel’. Proc of Reliability of the Ultrasonic Inspection of Austenitic Materials,
Belgium, 1980.
8. Morrell R., Ford D. A. and Harris K., ‘Calculation of Modulus in Different Directions for Single-
Crystal Alloys’. NPL Measurement Note DEPC (MN) 004.
9. Dandekar D. P. and Martin A. G., ‘Single crystal elastic constants of two nickel based superalloys’.
Journal of Materials Science Letters Vol 8, pp 1172-1173, 1989.
10. Dattatraya P., Dandekar P. and Martin A. G., ‘Temperature dependence of flexural and torsional
elastic constants of two nickel-based superalloys’. Journal of Materials Science Vol 25, pp 3321-
3326, 1990.
11. Whittle A. C., ‘Preliminary steps to validate a beam model for ultrasonic phased arrays’. Insight Vol
48, No 4, 2006.
12. Nageswaran C., Schneider C. R. A. and Bird C. R., ‘Sound field modelling using SimulUS’. Insight
Vol 50, No 5, 2008.
13. Calmon S., Mahaut S., Chatillon S. and Raillon R., ‘CIVA: An expertise platform for simulation and
processing NDT data’. Ultrasonics, Vol 44, pp 975-979, 2006.
14. Nageswaran C., Bird C. R. and Whittle A., ‘Immersion TRL phased array probe for stainless steel’.
Insight Vol 50, No 12, 2008.
15. Draft BS EN ISO 22825: Non-destructive testing of welds - Ultrasonic testing - Testing of welds in
austenitic steels and nickel-based alloys.
16. Draft BS EN ISO 13588: Non-destructive testing of welds - Ultrasonic testing - Use of (semi-)
automated phased array technology for examination of welds.
17. Nageswaran C., Carpentier C. and Tse Y.Y., ‘Improving phased array ultrasonic testing using
th
models to overcome austenitic weld distortion’. Proc of the 7 International Conference on NDE in
Relation to Structural Integrity for Nuclear and Pressurized Components, Japan, 2009.
18. Nageswaran C., Carpentier C. and Tse Y.Y., ‘Microstructural quantification, modelling and array
ultrasonics to improve the inspection of austenitic welds’. Insight Vol 51, No 12, 2009.
19. Carpentier C., Nageswaran C. and Tse Y.Y., ‘Evaluation of a new approach for the inspection of
th
austenitic dissimilar welds using ultrasonic phased array techniques’. Proc of the 10 European
Conference on NDT, Russia, 2010.
20. Schneider C.R.A, Kleiner D. and Williams S.M., ‘Optimising focussed/phased array ultrasonic
testing of clad pipes: Initial verification of CIVA 3D modelling software’. TWI Members Report
836/2005.
21. Nageswaran C., Schneider C.R.A. and Decourcelle N., ‘Validation of theoretical models for
focused/phased array ultrasonic inspection’. TWI Members Report 896/2008.
22. Tweedie A., Murray V. and Hayward G., ‘Aperiodic and deterministic 2D phased array structures
for ultrasonic imaging’. IEEE Ultrasonics Symposium, pp. 406 - 409, 2009.
23. Weston M., ‘Development of ultrasonic full matrix capture technology to improve flaw
characterisation’. TWI Members Report 966/2010.
24. Harvey G., Tweedie A., Carpentier C. and Reynolds P., ‘Finite element analysis of ultrasonic
phased array inspections on anisotropic welds’. Proc of the Review of Progress in Quantitative
Nondestructive Evaluation, USA, 2010.

Ref: TP AA058J-FR
Version: 01.05
Appendix A
Ultrasonic inspection of an austenitic weld - a case study

Ref: TP AA058J-FR
Version: 01.05
A1 Specimen
The specimen is representative of the surge end nozzle shown in Figure A1.
Figure A1 Diagram showing the geometry and dimensions of the surge end dissimilar weld
configuration (Copyright to British Energy Generation Ltd).
Figure A2 shows an image of the specimen (T1509) on the side with 2 vertical slots
(approximately 40mm deep) embedded along the centre line of the weld. In addition a
reference slot approximately 40mm deep is placed in the stainless steel parent, near the
edge of the block. The reference slot is inclined by approximately 10.3° to the vertical

Ref: TP AA058J-FR
Version: 01.05
Figure A2 Image of specimen T1509 showing two vertical slots along the centreline of the
weld and a reference slot in the stainless parent near the edge of the block.
Figure A3 shows the dimensions of the block, position of the slots and geometry of the
dissimilar weld. On the side opposite to the one with the slots three small porosity type flaws
are visible on the surface, as shown in the macrograph of Figure A4, and positional
information is given in Figure A5. The total number of micro-porosity flaws located under the
optical microscope number more than 6, as well as the three visible to naked eye.

Ref: TP AA058J-FR
Version: 01.05
Figure A3 Dimensions of specimen T1509 on the side with slots; slot near the cap is termed
slot H and the other as slot L, along with the reference slot.

Ref: TP AA058J-FR
Version: 01.05
Figure A4 Macrograph of the side containing the three sub-millimetre porosity type flaws
visible to the naked eye, with flaw A magnified.
Figure A5 Positional diagram of the side containing the porosity type flaws, shown in a
Cartesian coordinate system with the origin at the top right corner (ferritic parent).

Ref: TP AA058J-FR
Version: 01.05
A2 Approach
The aim of the study is to detect and position the slots within the weld body using the TRL
probe [A1]. Calibrations are performed on a ferritic calibration block with target 3mm side
drilled holes (SDH) at depths of 7, 18, 29 and 39mm from the top surface. In this work a
longitudinal velocity of 5900m/s is assumed (ferritic) and an attempt to correct for the
stainless steel is made through the use of the wedge delay trim.
A3 Results and discussion
The results are presented as sector scans on a dimensionally accurate weld overlay to aid
interpretation.
A3.1 Calibration scans
Figures A6 to A7 show the calibration scans on the 3mm side drilled holes (SDHs) along with
the gain setting to place the target signal at 80% of full screen height (FSH). In the present
work, the 45˚ beam is used for detection, with all signal-to-noise (S/N) ratio values measured
along this beam.
Figure A6 Calibration SDH at a depth of 7mm, measured at a gain of 38.5dB when target
signal is at 80% FSH.

Ref: TP AA058J-FR
Version: 01.05

Ref: TP AA058J-FR
Version: 01.05
Figure A9 Calibration SDH at a depth of 39mm, measured at a gain of 34dB when target
A3.2 Scan from stainless steel parent
Figure A10 shows the diffracted signal from the tip of the reference slot along the 45˚ beam
line; the focal law is set to focus at a depth of 6mm, with the wedge delay trimmed by +0.9μs
to place the echo at correct depth. The gain to place the signal at 80% FSH is 45.5dB, which
is 7dB higher than calibration (see Figure A6). The beam is highly focused at depths near the
surface; hence with a very tight field depth the beam is not sensitivity to other features of the
component. Note the faint echoes from the vertical side wall of the component.
Figure A11 shows the diffracted signal from the root of the reference slot along the 45˚ beam
line; the focal law is set to focus at a depth of 17mm, with the wedge delay trimmed by
+0.9μs, as for the tip echo (Figure A10). The gain at 80% FSH is 46dB, similar to that of the
tip echo. Figure A12 shows the diffracted signal from the root of the reference slot but with no
delay trim, ie 0μs. Note that the measured depth of the root signal is 15.3mm when trimmed
and 17mm with no trim. The actual measured position of the slot root is at 17mm (see Figure
A3). Therefore, the trim applied to correct for the tip depth position (as shown in Figure A10
where the tip echo is correctly measured at a depth of 6mm) leads to an error of 1.7mm in
the depth of the root echo. The delay trim value is a global correction that affects the
‘system’; hence, the error is due primarily to the inaccurate material velocity input.
However, both Figures A11 and A12 show the corner echo and the back wall echo from the
specimen; they are better positioned with respect to the material without the trimmed delay
(Figure A12) than when the delay is trimmed with the value used to correct for the tip echo
position.
All subsequent data presented in the report do not trim the wedge delay in order to better
understand the mechanisms involved in the positioning of echoes within the material.

Ref: TP AA058J-FR
Version: 01.05
Figure A10 Diffracted signal from the tip of the reference slot along the 45˚ beam line.
Figure A11 Diffracted signal from the root of the reference slot along the 45˚ beam line.

Ref: TP AA058J-FR
Version: 01.05
Figure A12 Diffracted signal from the root of the reference slot, without trim of the delay.
Figures A13 and A14 show detection of slot H with the probe focusing at depths of 8 and
18mm, respectively, where the weld overlay is dimensionally accurate.
Figure A13 Diffracted signal from the tip of slot H along the 45˚ beam; the gain is set to
50.25dB for 80% FSH.

Ref: TP AA058J-FR
Version: 01.05
Figure A14 Diffracted signal from the root of slot H along the 45˚ beam; the gain is set to
53.5dB for 80% FSH.
The gain settings to place the tip and root diffracted echoes at 80% FSH height are 11.75dB
and 22.75dB greater the corresponding calibration values, respectively. This illustrates firstly
the weakness of diffracted signals with respect to SDHs at the same range, but also the
greater scattering attenuation of the coarse grained material. Figure A15 shows the
measured A-scan along the 45˚ beam line, with the measured S/N greater than 6dB.
Figure A15 Measured A-scan along the 45˚ beam for the root diffracted signal from slot H.
In Figure A14 the probe is programmed to focus at the depth of slot H root, but a strong
signal is received from the tip of slot L. This could be due to beam skewing, where the actual
direction of sound energy is not propagating to the intended depth/position within the
material. However, when focused at the depth of the tip of slot L as shown in Figure A16, the
root of slot H is not illuminated strongly. This is an illustration of local anisotropic material
conditions strongly influencing beam forming within the weld volume. In addition, the gain is
15dB above the calibration setting, whereas it was 22.75dB to detect root of slot H at a lower
depth.

Ref: TP AA058J-FR
Version: 01.05
Figure A16 Diffracted signal from the tip of slot L along the 45˚ beam; the gain is set to
46.5dB for 80% FSH.
Figure A17 Diffracted signal from the root of slot L along the 45˚ beam; the gain is set to
49.5dB for 80% FSH.
In Figure A17, focusing at the root of slot L it is also possible to detect signals from the weld
root due to acoustic impedance mismatch leading to a corner effect type echo. The

Ref: TP AA058J-FR
Version: 01.05
measured depth of the root signal in Figure A17 is 34.3mm (the actual depth is 35.8mm); the
measured depth of the tip signal while focusing at the root is 24mm, giving a slot height of
10.3mm. The actual slot height was also measured at 10.3mm, possibly implying that error in
through-wall sizing may not be significant [A1].
The focal law was changed to a ‘distance’ type, where the sound is focused at a distance
ahead of the probe index point. Figure A18 shows probe focusing along a vertical line
corresponding to the centreline of the weld where the 45˚ beam impacts on the tip of slot L.
The beam strength along the 45˚ beam line is stronger than at beam angles corresponding to
the root of slot L, as well as the tip and root of slot H. This has implications towards the
design of inspection procedures.
Figure A18 Focusing along a vertical line corresponding to the centreline of the weld where
the 45˚ beam is incident on the tip of slot L.
Figure A19 shows detection of a possible porosity point within the weld volume, as shown in
Figure A4, because the signal does not have any lateral length within the material. Note also
the error in through-wall positioning of the diffracted signals. The law was set to focus at a
depth of 22m where the porosity signal was identified, but positioning of legitimate echoes
have significant errors.

Ref: TP AA058J-FR
Version: 01.05
Figure A19 Detection of a point signal that suggests the presence of porosity.
A3.3 Scan through buttering from the ferritic parent
Figures A20 and A21 show the scans for the tip and root diffractions of slot H, with the law
focused at the corresponding depths. As before, the tip signal from slot L is stronger than the
root signal from slot L when focused at the depth level of the root of slot L (Figure A21),
implying the presence of distortive effects.
Figures A22 and A23 show the tip and root diffractions from the tip and root of slot L, with the
law focused at the corresponding depths. The gain settings for detection of the tip and root
signals are 10.5dB and 18.25dB above calibration gains at the corresponding depths. Figure
A23 also shows that severe beam distortion could be taking place when focused at the depth
of the root of slot L, with strong echoes from the weld root and back wall.

Ref: TP AA058J-FR
Version: 01.05
Figure A20 Diffracted signal from the tip of slot H along the 45˚ beam; the gain is set to 48dB
for 80% FSH.
Figure A21 Diffracted signal from the root of slot H along the 45˚ beam; the gain is set to
52dB for 80% FSH.

Ref: TP AA058J-FR
Version: 01.05
Figure A22 Diffracted signal from the tip of slot L along the 45˚ beam; the gain is set to 42dB
for 80% FSH.
Figure A23 Diffracted signal from the root of slot L along the 45˚ beam; the gain is set to
52.25dB for 80% FSH.

Ref: TP AA058J-FR
Version: 01.05
A4 Recommendations
1. The work was performed with the nominal velocity set to that of ferritic steel
(5900m/s). Positioning accuracy is greatly dependant on knowing the actual velocity
of the sound wave along the propagation direction; hence, an essential task will be to
measure the velocity of the stainless steel parent at different beam angles. Test
blocks of the actual stainless steel material (or similar) need to be sourced, with
SDHs at various ranges.
2. The focal law driving the probe has significant influence on the regions where
accurate positioning can be performed - ie regions where significant errors could be
induced must be defined in any procedure.
3. Study the effect of the material as the probe is moved along the weld, in order to
gauge how significantly the weld microstructure changes along the weld.
A5 References
A1. Nageswaran C., Bird C. R. and Whittle A., ‘Immersion TRL phased array probe for
stainless steel’. Insight Vol 50, No 12, 2008.

Ref: TP AA058J-FR
Version: 01.05
Appendix B
Specification of the TRL-1 array

Ref: TP AA058J-FR
Version: 01.05
B1 Description
This document is the specification of the phased array probe TRL-1 generated after iteration
3 of the first level specification study (see Section 4.4.1).
B2 Ultrasonic parameters
Parameter Value Comments

Probe type Transmit – Receive
Generated wave type Longitudinal
Array type 2D matrix
Number of elements on primary
32 See diagram
axis (np)
Element pitch of primary axis (pp) 4mm See diagram
Number of elements on secondary
2 See diagram
axis (ns)
Element pitch of secondary axis
6mm See diagram
(ps)
Inter element spacing 0.1mm
Frequency 1.5MHz
Bandwidth (-6dB) 80%*
Pulse length (-20dB) 1.69µs*
Wedge material Water
Wedge angle (i) 13.1º See diagram
Wedge surface N/A
Wedge condition N/A
Roof angle (r) 2.5º See diagram
Distance between the two inside
11mm See diagram
edges of the array (l)
1 dB between
Homogeneity in sensitivity
adjacent elements
Inter element cross talk < -40dB
<-60dB relative to a
Parasitic / internal echoes
flat target echo*
*measured in the conditions below
B3 Physical and electrical parameters

Dependent on holding
Overall length ~150mm
frame
Dependent on holding
Overall width ~100mm
frame
Overall height (with cable
>100mm
connector)
Cable 50 multi-coaxial with

overall shielding

Ref: TP AA058J-FR
Version: 01.05
Cable output On top, centred plus

cable restraint
Cable length 5m Halogen free jacket
Cable test Simulated impedance
at the end of the cable
to be calculated
Electrical matching None
Only materials approved
Housing material Stainless steel for use in a Reactor
Pressure Vessel
Hypertronics
Connector type compatible with R/D-
Tech FOCUS
Water coupling system N/A
B4 Identification/engraving
DISSIMILAR
TRL 1.5 MHz – 32x2 Elts
4mm Pitch Primary Axis
6mm Pitch Secondary Axis
Elements No 1 and 32 positions to be shown
B5 Using conditions*

Chemical constraints Contact with water or
ultrasonic water
based gel
Temperature 0° to 40° (storage)
10 to 40° (functioning)
Excitation Negative pulse 200V
maximum (measured
in 50 environment)
Max Repetition Frequency 10kHz
Shocks and Vibrations Operation on a
submersed robotic
delivery system
Nuclear Radiation None
Normative Constraints None
B6 Pre-delivery checks
Impedance measurement for all the elements at the nominal frequency.

Inter-element coupling measurement on 2 pairs of elements.
Checks performed in water on a flat target in pulse echo mode at 25mm to 100mm
distance.
Excitation signal: 1 negative pulse.
Sensitivity homogeneity for all the elements.
Temporal response & frequency spectrum measurement for 8 elements.

Ref: TP AA058J-FR
Version: 01.05
B7 Documentation to be supplied
Specifications of the probe

Overall drawing of the probe
Measurement results
B8 Guarantee
One year guarantee against design and manufacturing defects.
B9 Diagram
Figure B1: TRL-1 probe profile along the Figure B2: TRL-1 probe profile along the
primary axis secondary axis
Tx: Transmit array.

Rx: Receive array.
Lp: Aperture along the primary axis (np x pp).

Ls: Aperture along the secondary axis (ns x ps)
i: Wedge angle.
r: Roof angle.
l: Distance between the two inside edges of the array.
np: Number of element along the primary axis.

ns: Number of element along the secondary axis.
pp: Element pitch along the primary axis.

ps: Element pitch along the secondary axis.

Ref: TP AA058J-FR
Version: 01.05
Appendix C
Specification of the 2D-1 array

Ref: TP AA058J-FR
Version: 01.05
C1 Description
This document is the specification of the array probe 2D-1 generated from first level
specification study (see Section 4.5).
C2 Ultrasonic parameters

Probe type 2D Array
Generated wave type Longitudinal
Array type 2D matrix
Number of elements on primary
18
axis (np)
Element pitch of primary axis (pp) 3.5mm
Number of elements on secondary
7
axis (ns)
Element pitch of secondary axis
6.4mm
(ps)
Inter element spacing 0.1mm
Frequency 1.5MHz
Bandwidth (-6dB) 80%*
Pulse length (-20dB) 1.69µs*
Wedge material Rexolite
Wedge angle (i) 18.5 See diagram
Wedge surface N/A
Wedge condition N/A
Homogeneity in sensitivity 1 dB between
adjacent elements
Inter element cross talk < -40dB
Parasitic / internal echoes <-60dB relative to a
flat target echo*
*measured in the conditions below
C3 Physical and electrical parameters

Overall Length 96mm
Overall Width 96mm
Overall Height (without cable
54mm
connector)
Cable 50 multi-coaxial with
overall shielding
On top, centred plus
Cable Output
cable restraint
Cable Length 5m Halogen Free Jacket
Cable Test Simulated impedance
at the end of the cable
to be calculated

Ref: TP AA058J-FR
Version: 01.05
Electrical Matching None

Only materials approved
Housing Material Stainless steel for use in a Reactor
Pressure Vessel
Hypertronics
Connector Type compatible with R/D-
Tech FOCUS
Water Coupling System N/A
C4 Identification/engraving
DISSIMILAR
2D Array 1.5 MHz – 18x7 Elts
3.5mm Pitch Primary Axis
6.4mm Pitch Secondary Axis
Elements No 1 and 18 positions to be shown
C5 Using conditions*

Chemical constraints Contact with water or
ultrasonic water
based gel
Temperature 0° to 40° (storage)
10 to 40° (functioning)
Excitation Negative pulse 200V
maximum (measured
in 50 environment)
Max Repetition Frequency 10 kHz
Shocks and Vibrations Operation on a
submersed robotic
delivery system
Nuclear Radiation None
Normative Constraints None
C6 Pre-delivery checks
Impedance measurement for all the elements at the nominal frequency.

Inter-element coupling measurement on 2 pairs of elements.
Checks performed in water on a flat target in pulse echo mode at 25mm to 100mm
distance.
Excitation signal: 1 negative pulse.
Sensitivity homogeneity for all the elements.
Temporal response & frequency spectrum measurement for 8 elements.
C7 Documentation to be supplied
Specifications of the probe

Overall drawing of the probe
Measurement results

Ref: TP AA058J-FR
Version: 01.05
C8 Guarantee
One year guarantee against design and manufacturing defects.
C9 Diagrams
Figure C1: 2D-1 array probe profile along Figure C2: 2D-1 array probe profile along the
the primary axis and secondary axis primary axis

Ref: TP AA058J-FR
Version: 01.05
Appendix D
Specification of the MicroPulse 5PA

Ref: TP AA058J-FR
Version: 01.05
Product overview
Phased Array MicroPulse (64/64, 128/128, 256/256, 512/512, all channels may be used for beam
forming). The unit has the flexibility to output individual A-scans prior to summation and optional
separate channels for high-performance pulse-echo and TOFD (available in multiples of 16 channels).
Software platforms
PNL ArrayGen with SimulUS beam modelling software as standard and compatible with British Energy
MIPS/GUIDE and UTEX Winspect / InspectionWare. Open data format and long-established
MicroPulse command language mean that the users have the option to write their own applications.
Contact details
Peak NDT Ltd

Unit 1, Enterprise Way,
Jubilee Business Park
Derby
DE21 4BB
Tel: +44(0)1332 738752 fax: +44(0)1332 73887 e-mail: sales@peakndt.com
NOTE: Peak NDT Ltd. reserves the right to change these specifications without notice.

Ref: TP AA058J-FR
Version: 01.05
Specification of phased array channels
Parameter Range Step Size

Pulser Type Negative square wave N/A
Pulser Voltage 5 to 200Volts 5Volt
Pulser Rise Time <5ns N/A

Pulser
Pulser Width 20nsec to 500nsec 2nsec
Pulse Repetition Frequency 1Hz to 20kHz 1Hz
Pulser Delay 0 to 25000nsec 1nsec
Gain 70dB 0.25dB
Input Noise 2nV typical N/A
Gain Linearity Better than 0.25dB N/A
Input Impedance 50Ω N/A
Bandwidth 0.75MHz to 20MHz (-3dB) N/A

Receiver 0.75MHz to 5MHz (-3dB) Bandpass Filter
5MHz to 10MHz (-3dB) Bandpass Filter
Filters Discrete selection
0.75 to 20MHZ Broadband
Receiver Delay 0 to 25000ns 1nsec
Dynamic Depth Focusing At 100MHz realtime N/A
Channel Crosstalk Better than 60dB between channels at 2MHz N/A
DAC Dynamic Range 0 to 40dB 0.25dB
DAC Trigger Transmit pulse or material interface echo User selectable

Distance
Amplitude No of DAC curves 256 utilising up to 64kbytes N/A
Correction DAC update 40dB/µsec N/A

0.78125MHz, 1.5625MHz, 3.125MHz, 6.25MHz,
DAC clock rate
12.5MHz and 25MHz selectable
ADC Resolution 12 bits N/A
ADC Rate 10, 25, 50 and 100MHz
Number of ADC’s One per channel
Element Summing Up to 512 channels N/A

No Rectification
Fullwave
Rectification Discrete selection
+ve halfwave
Digitiser -ve halfwave
and Digital Post Rectification Filter None and 7 selectable settings
Processing Gates 1 gate of up to 32k sample points N/A
Gate Delay 64k sample points from trigger or I/F echo

for each gate up to 80 peaks (N + largest), first
Hardware Peak Processing
peak, largest peak
Peak Threshold 5 to 2047% ½%
Averaging 2 to 256 realtime
GRE 1 element, n elements or summed waveform

Ref: TP AA058J-FR
Version: 01.05
Specification of conventional channels
Parameter Range Step Size

Pulser Type Negative square wave N/A
Pulser Voltage 50 to 300Volts 50Volt
Pulser Rise Time <5ns N/A

Pulser
Pulser Width 20nsec to 500nsec 2nsec
Pulser Damping 50Ω to 660Ω in 8 steps N/A
Pulse Repetition Frequency 1Hz to 20kHz 1Hz
Gain 70dB 0.25dB
Input Noise 2nV typical N/A
Gain Linearity Better than 0.25dB N/A
Input Impedance 660Ω N/A
Bandwidth 0.75MHz to 25MHz (-3dB) N/A

0.75MHz to 12MHz (-3dB) Bandpass Filter
2.5MHz to 18MHz (-3dB) Bandpass Filter
Receiver 3MHz to 22MHz (-3dB) Bandpass Filter
0.5MHz Bandpass Filter
1MHz Bandpass Filter
Filters 2MHz Bandpass Filter Discrete selection
5MHz 2nd order TOFD Bandpass Filter
10MHz 2nd Order TOFD Bandpass Filter
Channel Crosstalk < 60dB between channels at 2MHz
DAC Dynamic Range 0 to 40dB 0.25dB
DAC Trigger Transmit pulse or material interface echo User selectable

Distance
Amplitude No of DAC curves 32 utilising up to 32kbytes N/A
Correction DAC update 40dB/µsec N/A

0.78125MHz, 1.5625MHz, 3.125MHz, 6.25MHz,
DAC clock rate
12.5MHz and 25MHz selectable
ADC Resolution 12 bits N/A
ADC Rate 10, 25, 50 and 100MHz N/A

No Rectification
Fullwave
Rectification Discrete selection
+ve halfwave
-ve halfwave
Digitiser Post Rectification Filter None and 7 selectable settings N/A
and Digital
Gates 1 gate of up to 32k sample points
Processing
Gate Delay 64k sample points from trigger or I/F echo
for each gate up to 80 peaks (N + largest), first
Hardware Peak Processing
peak, largest peak
Peak Threshold 5 to 2047% ½%
Averaging 2 to 256 realtime

Ref: TP AA058J-FR
Version: 01.05
General specifications
Phased Array Connector 160-pin female connector. Hypertronics™ HLMYJPAPF1600
Conventional UT Connector Triaxial 1S connector. Lemo ERA.1S.650.CTL
Interface Gigabit Ethernet capable of up to 20Mbyte per second

Connectors 4 axes of 32 bit encoder inputs accepting encoders between 5 and 15Volt
Encoders
and at rates of up to 700kHz
and Digital I/O
8TTL compatible inputs and 8 open collector outputs capable of sinking up to
400mA
Interfaces Trigger, Gate, A-scan. Showing either an individual channel or a summed
Oscilloscope Outputs
etc waveform (reconstituted analogue signal obtained from digitised waveform)
Case Size 450mm x 380mm x 170mm
Power Supply 90-260 VAC at 45-100Hz
Weight 15kgs

Ref: TP AA058J-FR
Version: 01.05
Appendix E
Automated baseline inspection report (British Energy)

NOT PROTECTIVELY MARKED E/TSK/GEN/7228/DISSIMILAR19
Revision 000 Phased array inspection of austenitic welds
Page 202 of 242
Engineering
Report
TSB Sponsored Project:

Phased Array Ultrasonic Inspection of
Dissimilar Metal Welds
MIPS / GUIDE Automated Ultrasonic Inspection

of 'DISSIMILAR' Project Test Welds
- ‘Conventional’ Fingerprint Inspection
By: JE Pearce, IG Lincoln Date: July 2009
British Energy Generation Limited

Central Engineering Support
Engineering Technology Branch
Inspection Group
NOT PROTECTIVELY MARKED Document template Rev09.001

NOT PROTECTIVELY MARKED E/TSK/GEN/7228/19
Revision 000
Page 203 of 242
BRITISH ENERGY GENERATION LIMITED

CENTRAL ENGINEERING SUPPORT
© 2009 Published in the United Kingdom by British Energy Generation Ltd.
All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means,
including photocopying and recording, without the written permission of the copyright holder, BEGL,
application for which should be addressed to the publisher. Such written permission must also be obtained
before any part of this publication is stored in a retrieval system of any nature.
LIMITATION OF LIABILITY – Whilst BEGL believes that the information given in this document is correct at
the date of publication it does not guarantee that this is so, nor that the information is suitable for any
particular purpose. Users must therefore satisfy themselves as to the suitability of the information for the
purpose for which they require it and must make all checks they deem necessary to verify the accuracy
thereof. BEGL shall not be liable for any loss or damage (except for death or personal injury caused by
negligence) arising from any use to which the information is put.

Revision 000
Page 204 of 242
SUMMARY SHEET
Project: TSB Sponsored Project: Phased Array Ultrasonic Inspection of
Dissimilar Metal Welds
Document Title: MIPS / GUIDE Automated Ultrasonic Inspection of 'DISSIMILAR'

Project Test Welds – ‘Conventional’ Fingerprint Inspection
Document Reference: E/TSK/GEN/7228/19
Document Author: John Pearce, Ian Lincoln
Date: July 2009
Task File Number: E/TSK/GEN/7228
Summary
Test specimens comprising 85 mm thick Inconel safe-end transition welds have been produced as part of
the TSB sponsored DISSIMILAR Project. As a contribution to the project, BE Inspection Group have
performed an automated ultrasonic inspection of the two blocks, using the MIPS / GUIDE system. This
was to provide a ‘conventional’ inspection fingerprint of the welds and the implanted defects, and to
provide a baseline of current inspection capability for subsequent comparison with optimised Phased
Array ultrasonic inspections.
The two blocks have been scanned from both outer and inner surfaces and both axial directions, using a
range of 0° and angled compression wave beams. Additional sizing scans have been implemented to
provide the best sizing information possible.
The inspection data has been interpreted by two experienced data interpreters and the findings and
measurements have been verified before being tabulated and presented in this report. The results take
the form of signal-to-noise ratio values for the defects detected and best estimate positioning and sizing
values.
The usual guidelines for effective inspection of thick-section austenitic welds are borne out: use all
available scanning surfaces, attempt to minimise beam paths through weld metal, use beams with
appropriate arrangements to maintain effective beam shapes, match inspection frequencies to materials,
etc..
The capability demonstrated on these test specimens is consistent with BE experience from inspection
qualification exercises on similar welds.
Conclusions
(1) The two test blocks produced under the DISSIMILAR Project have been inspected using BE’s
MIPS / GUIDE automated ultrasonic inspection system and using methods and techniques typical
of the inspection of this type of joint in nuclear power plant.
(2) The data produced by these inspections provides a ‘conventional’ fingerprint of the welds, for future
comparison with optimised Phased Array inspections.
(3) All the defects have been detected, with varying margins of signal-to-noise ratio.
(4) The defects have been positioned and sized within the expected tolerances for this thickness of
austenitic weld.
(5) The results of all the defect detection assessments and positioning and sizing measurements have
been 100% verified and tabulated in this report.
rd
(6) The delivery of this report completes BE’s Action 11 form the 3 Quarterly Meeting.

Revision 000
Page 205 of 242
VERIFICATION CERTIFICATE
Document Title: MIPS / GUIDE Automated Ultrasonic Inspection of
'DISSIMILAR' Project Test Welds – ‘Conventional’ Fingerprint
Inspection
Project: TSB Sponsored Project: Phased Array Ultrasonic Inspection
of Dissimilar Metal Welds
Document Reference: E/TSK/GEN/7228/19
Revision: 000
Task File Number: E/TSK/GEN/7228
QA Grade of document: 4
Author: John Pearce, Ian Lincoln
Section/Group: Inspection Group / Engineering Technology Branch
Verification Applied: i) Review of MIPS Setups and inspection records
ii) Separate and independent interpretation of data
iii) Verification of Detection and Signal-to-Noise Ratio Tables
iv) Verification of Defect Position and Sizing Tables
v) Review of Report, checks for accuracy, Figures, Tables
Technical Reviewer(s): Not applicable. Report produced internally.
Section/Group:
Technical Review
Details:

Revision 000
Page 206 of 242
DOCUMENT ISSUE / AMENDMENT CERTIFICATE
Date Authors Revision Amendment
July 2009 JE Pearce, IG Lincoln 000 Original Issue

Revision 000
Page 207 of 242
GLOSSARY
BWE Backwall Echo (strong planar response from far surface, usually from 0° beam)
BWR Boiling Water Reactor (nuclear plant type)
DAC Distance-Amplitude Correction (increasing gain with range along beam)
DMW Dissimilar Metal Weld
EDP Echodynamic Pattern (characteristic signal response)
EMI Electromagnetic Interference (noise in ultrasonic signals from external source)
FSH Full-Screen Height (100% deflection measurement on flaw detector instrument)
GUIDE Graphical Ultrasonic Inspection Data Evaluation – BE data analysis system
MMA Manual Metal Arc (welding process)
MIPS Micropulse Inspection and Processing System – BE data acquisition system
MRF MIPS Results File (initial output from MIPS – raw data file)
PRF Pulse Repetition Frequency (control of time gap between probe firings)
PRG Primary Reference Gain (gain to bring Reference Reflector signal to 80% FSH)
PWR Pressurised Water Reactor (nuclear plant type)
RGC Run Gain Correction (additional gain added to modify inspection sensitivity)
RPV Reactor Pressure Vessel (central part of nuclear power plant)
SDH Side-Drilled Hole
SNR Signal-to-Noise Ratio (discrimination between signal and background noise)
TRL Transmit-Receive Longitudinal (ultrasonic compression wave probe design)
TSB Technology Strategy Board

Revision 000
Page 208 of 242
CONTENTS
SECTION TITLE PAGE
1. Purpose 209
2. Background 209
3. Inspection Techniques and Procedure 210
4. Physical Setup and Equipment 211
5. Scanning Parameters and MIPS Setup 212
6. Description of Data Files Produced 214
7. Detection Results 215
7.1 Features of Data 215

(geometric echoes, parent materials, inclusions, attenuation)
7.2 Signal-to-Noise Ratio (SNR) Tables 215
7.3 Penetration of Weld 216
7.4 Analysis and Discussion 216
8. Positioning and Sizing Results 217
8.1 Additional Sizing Scans 217

8.2 Defect Positions 217
8.3 Best Estimates Defect Sizes 218
8.4 Analysis and Discussion 218
9. Review 219
10. Conclusions 219
11. References 220
Figures 221
Tables 231
Distribution / Notification List 242

Revision 000
Page 209 of 242
1. Purpose
This report summarises the activities carried out at British Energy, Barnwood, in May and June
2009, in connection with the TSB sponsored ‘DISSIMILAR’ project on Phased Array Ultrasonic
Inspection of Dissimilar Metal Welds. Specifically, the report describes the ‘conventional’
automated ultrasonic inspection of the 85 mm thick Inconel safe-end transition weld specimen
produced under the ‘DISSIMILAR’ project. The two parts of this specimen were subject to
detailed inspection using the British Energy MIPS / GUIDE system, to provide a ‘fingerprint’ of
the welds and the implanted defects, and to provide a baseline of ‘conventional’ inspection
capability for comparison with subsequent Phased Array inspections.
2. Background
The TSB sponsored ‘DISSIMILAR’ project has the main aim of using Phased Array technology
to improve the reliability of inspection of safe-end transition welds (or Dissimilar Metal Welds,
DMW, in American parlance). This is to be achieved by:
i) better knowledge of austenitic weld metal grain alignment, alignment distributions and
alignment variations;
ii) advanced modelling of sound wave propagation through textured and anisotropic weld
metal;
iii) advanced design and prototyping of new Phased Array transducers, employing computer
modelling and improved piezo-electric materials;
iv) high-specification ultrasonic instrumentation, capable of handling high element number

arrays;
v) formulation and testing of advanced delay law algorithms to mitigate for the disruptive
effects of propagation through textured and anisotropic weld metal.
As part of this project, a heavy-section safe-end transition weld specimen has been produced,
to provide test material to assess weld metal grain alignment and homogeneity, and to test the
new inspection equipment and techniques. This weld is 85 mm thick and the original block was
more than 600 mm long. It was welded with Inconel 182 filler weld metal, using the Manual
Metal Arc (MMA) process and it is generally typical of safe-end transition welds between low
alloy steel vessels and stainless steel pipework in light water-cooled nuclear plant (PWR, BWR).
Figures 1A and 1B illustrate the dimensions and defect details of the specimen, which was
sectioned into two parts, designated Block ‘A’ and Block ‘B’ (Figure 2).
As part of their contribution to the project, British Energy offered to perform a ‘conventional’
automated ultrasonic inspection of the weld specimen, using equipment and techniques typical
of those currently employed to inspect such welds in nuclear plant. This would involve the use
of the MIPS / GUIDE system and be intended to provide the following:
i) a ‘fingerprint’ record of the test specimen weld and the implanted defects;
ii) a baseline of typical inspection capability using ‘conventional’ techniques and equipment,
for comparison with subsequent Phased Array inspections.

Revision 000
Page 210 of 242
3. Inspection Techniques and Procedure
Due to the specimen’s similar thickness to the Sizewell B Steam Generator Primary Nozzle
safe-end transition weld, the probe set for that inspection was chosen as the basis for the
techniques (11.1), but with a number of additional probes to provide the optimum capability. The
full list of probes / beams is given below in Table 1, but they include:
i) single- and twin-crystal 0° compression;

ii) twin-crystal angled compression (TRLs) at 1, 1.5 and 2 MHz;
iii) single-crystal (elliptical) angled compression.
Since the test specimen weld also resembles typical PWR RPV nozzle safe-end transition welds
(older US, Westinghouse-design plant), it was decided that inspection from the internal surfaces
would also be performed, as that is typically the access for RPV weld inspection.
Scanning would use a XY frame, with the block immersed in water in a custom-made stainless
steel container. Scans would be performed from both external and internal surfaces, and in both
axial directions. No transverse scans would be performed.
Scanning increments of 2 mm Primary by 4 mm Secondary would be used for search scans

(typical for this size of weld). This means that the probes would be fired at every 2 mm pitch
along the scanning direction (Primary line) and then the next Primary line would be scanned
after a 4 mm sideways (Secondary) step.
Scanning sensitivity would be set to collect material noise (‘grass’) on all scans, but with the
calibration and scanning sensitivity recorded using Ø3 mm Side-Drilled Holes (SDHs) in fine-
grain, low-attenuation forged austenitic calibration blocks. MIPS Distance-Amplitude Correction
(DAC) would be employed on all scans.
MIPS processing would allow the data from all scans to be plotted relative to the same
component co-ordinate system and datum reference points.
Data interpretation would be based on thresholding the data to an amplitude corresponding to

the peak grass level, and then marking clusters of indications with Signal-to-Noise Ratios
(SNRs) of 6 dB or more. These would then be boxed and collated using the ‘Box / Crate’
facilities within GUIDE.
Once the main defect areas have been identified, then more detailed sizing scans, using
smaller scan increments, and some different probes / beams, would be applied to maximise the
sizing and characterisation information.
Defect positioning and sizing are performed using methods corresponding to the echodynamic
patterns (EDPs) exhibited by the defects. Generally, Pattern 2 EDPs in the longitudinal direction
indicate a smooth reflector with length, and positioning and sizing are done by the 6 dB
amplitude drop method. Pattern 2 EDPs in the throughwall direction indicate a planar reflector
with significant throughwall extent. These are also sized by the 6 dB drop method. Pattern 3
EDPs often originate from rough defects, and the maximum amplitude method is employed to
make estimates of position and extent. The most reliable method for throughwall sizing will be
by the detection, identification and positioning of diffracted edge wave signals from defect
extremities. Maximum amplitude sizing is used whenever defect tip echoes are visible. All of
these sizing aspects are made less straightforward by the characteristics of austenitic weld
metal.

Revision 000
Page 211 of 242
4. Physical Setup and Equipment
The inspection of the DISSIMILAR Blocks ‘A’ and ‘B’ used a normally-configured MIPS system:
MIPS PC with printer
Micropulse
Motor Controller
Manipulator
The manipulator used was an existing XY scanning frame with adjustable height feet and
modular probe holder fittings. This had sufficient Primary stroke and Secondary travel to cover
the scan lengths required. The Primary drive does suffer from a minor degree of backlash, but
was acceptable for search scans. For the sizing scans, all data collection was performed whilst
scanning in one direction, to eliminate the effects of backlash.
A specially-made stainless steel immersion tank was fabricated to allow the blocks to be
scanned immersed in water. The tank was situated on the floor to reduce the need to lift the
blocks and to minimise the height to which they would need to be lifted to position them, mainly
for handling safety reasons.
Figure 4 shows the setup of the XY frame and the immersion tank.
BE currently uses the Peak NDT Micropulse 4 generation of digital, programmable ultrasonic
flaw detectors as its mainstay for automated inspections using MIPS / GUIDE. BE has one
Micropulse 5 PA with a second on order.
Block ‘B’ was scanned first, using a Micropulse 4 (S10056). Block ‘A’ was scanned later, using a
Micropulse 5 PA (S15029) set to mimic a Micropulse 4. Both units had been calibrated within
the previous 12-month period, and both were subject to linearity checks (timebase and
amplifier) prior to use.
The probe set used for these inspections is detailed in Table 1 below. These are a combination
of single- and twin-crystal compression wave probes, mainly produced by one of the leading
manufacturers of such specialised probes, Applus-RTD, Rotterdam, the Netherlands. The
probes were all subject to functional performance checks prior to use.
In order to ‘harden’ the system against noise and electromagnetic interference (EMI), the
cabling between the Micropulse and the probes has no inter-connections and is made from
high-specification Sűhner TriAx cable. The connections at the Micropulse and probe ends are
made with Lemo TriAx connectors. The same leads used for scanning were used for the probe
checks and the calibration.
The calibration blocks used to check the probes and to set the probe shoe delays and the
scanning sensitivity were R25 and R50 radius blocks and a rectangular DAC block with a range
of Ø3 mm SDHs at different depths. These blocks are made from low-attenuation forged
austenitic stainless steel with a known fine grain size (ASTM 5 or better, by E112-96). These
blocks are detailed on the BE Inspection Group Test Block Register database. Figure 6
illustrate the 3 blocks used for this work. Due to the small differences in velocities between
Inconel and austenitic stainless steel, there may be small errors in positioning due to angle and
range changes, but there are three different materials in this weld, so a compromise choice
needs to be made.

Revision 000
Page 212 of 242
5. Scanning Parameters and MIPS Setup
All the scanning employed the same MIPS PC, running the V1.22 version of MIPS. Each scan
with a particular beam, in each of the four surface / direction combinations, is defined with a
separate MIPS Setup file. So-called ‘long’ MIPS codes were used to allow more meaningful
Setup naming:
Site Code: TWI
Component Code: BLA, BLB – Block ‘A’, Block ‘B’
Run Code: OD0XX – outer surface, sequential numbering

ID0XX – inner surface, sequential numbering
ODSXX – outer surface, sizing scan
IDSXX – inner surface, sizing scan
20 search scans from the outer surface and 16 search scans from the inner surface were
carried out on each block.
Figure 5 illustrates the XYZ co-ordinate system used.
The X axis is normal to the weld and +X runs towards the austenitic side, -X runs towards the
ferritic side. X = 0 is the weld centreline, taken as 250 mm in +X from the ferritic end edge.
The Y axis is parallel to the weld. +Y runs left to right when looking down on the outer surface
towards +X. Y = 0 is at the left hand edge when looking down on the outer surface towards +X.
The Z axis is normal to the surface, with Z = 0 at the outer surface and +Z towards the inner
surface.
The use of the MIPS Processing AG2 Geometry Code allows all the data from all beams to be
imaged with reference to the same co-ordinate system and the same datum reference points,
regardless of scanning direction or surface. The system is given details of the change of surface
and the block thickness, and the geometry code sorts out all the plotting issues. Note that the
data is also fully corrected for beam angle and emission point offsets, and effectively plotted in
the real-space co-ordinates from which it originated.
Some of the common data entered for use across multiple scans:
Compression Velocity, VL = 5750 ms-1

Shear Velocity, VT = 3150 ms-1
(velocities for austenitic stainless steel – half the scans are through this material and the
calibration blocks are also made of it)
Micropulse Pulser Voltage = 300 V
Micropulse Digitisation Frequency = 25 MHz
Micropulse Peak Detection per Test Gate = 20 = Largest indication + next 19 from front of test
Because it was an inspection on austenitic materials, and high levels of material noise were
expected, as each probe was scanned, the associated MIPS Beam was divided into a number
of shorter test gates, to ensure that the system’s peak detection capacity was not exceeded.
Therefore a beam interrogating a range of 95 – 225 mm (130 mm) would have 6 overlapping
tests, none more than 30 mm long.
Revision 000
Page 213 of 242
As part of the probe performance checks, the MIPS Beam parameters for each probe are
optimised – these include excitation pulse width, receiver bandwidth / filter, signal smoothing
and damping.
Along with each search beam, an associated grass coupling check beam was implemented.
This monitors a set gate for a specified grass amplitude level, to ensure adequate coupling.
Whilst this is done routinely and represents good practice, when scanning in immersion, it really
only serves to indicate if the scanning surface is very poor or the probe has tripped over
something. No issues with coupling were encountered on any of the scans covered by this
report.
The probes were calibrated and the scanning sensitivity was set as follows:
i) a specific Ø3 mm SDH is chosen to be the Reference Reflector for each beam – usually a
hole at a depth corresponding either to the beam focus (TRLs) or at the start of the DAC;
the range of this reflector is checked and a gate bracketing this range is entered in MIPS
as the Full Calibration gate;
ii) the amplitude required to bring the Reference Reflector to 80% Full-Screen Height (FSH)
is the Primary Reference Gain or PRG. This is set as the MIPS Calibration Gain. A MIPS
Full Calibration routine demands that the signal from the Reference Reflector is acquired
in the calibration test gate and brought to between 70 and 90% FSH by adjustment of the
Calibration Gain; the results of this sequence is automatically printed out, as part of the
inspection records;
iii) with a data acquisition (recording) threshold of 10% FSH, then the applied sensitivity is Ø3
mm SDH + 18 dB; a minimum of 2 dB Run Gain Correction (RGC) is added to bring the
applied sensitivity up to Ø3 mm SDH + 20 dB or 10% DAC; this means that the lowest
amplitude signal which will be recorded will be one which is 20 dB weaker than that from a
Ø3 mm SDH;
iv) for inspections on austenitic materials, it is general policy to increase the sensitivity until
material noise is recorded. At this level, any defect with significant signal-to-noise ratio
above the material noise will definitely also be recorded. This usually occurs for
sensitivities of Ø3 mm SDH + 20-30 dB, depending on the parent materials and the
coarseness of the weld structure.
Clearly, for such inspections on austenitic materials, due to the effects of attenuation, the term
sensitivity does not have the same quantitative meaning as it does for other inspections, and
many measurements are made relative to peak noise levels. However, BE procedures require
all measurements to be referenced to the nominal sensitivity with respect to a Ø3 mm SDH
DAC. This helps with inspection repeatability and comparison between data sets.
For all the search scans, the scanning pattern used a 2 mm Primary increment and a 4 mm
Secondary increment, both typical for this size of weld. For sizing scans, these increments were
reduced to 0.5 x 0.5 mm, and data collection was performed in a single scanning direction to
mitigate against the slight backlash on the Primary drive.

Revision 000
Page 214 of 242
The target inspection volume to be covered by the range of search beams was defined as the
full volume of weld metal and parent material up to 15 mm from the weld fusion faces, as shown
in Figure 7. The various different beams are aimed to cover specific areas of this inspection
volume and no one beam covers the full volume, rather the coverage combines to provide full,
and usually redundant, coverage of the complete inspection volume.
6. Description of Data Files Produced
As noted above, each probe / beam scanned produces a single, separate MIPS Results file. So
with 36 search scans per block and 29 sizing scans in total, there were a total of 101 MIPS
Results files (.MRFs) comprising 196 MB.
When processed, each MRF produced 2 MIPS Flaw files, one (Beam 01) containing the
geometry-corrected data from the inspection beam and the second (Beam 02) the results of the
grass coupling monitoring. (The 0° beams did include a 2nd search beam to cover the Backwall
Echo (BWE) at reduced gain, so these 4 files produced 3 Flaw files each). Therefore, all
assessment work was performed on the Beam 01 files. There were 101 of these produced,
totalling 516 MB.
A GUIDE data interpretation PC, running the V1.13d version of GUIDE was then used to image
the Flaw files, initially producing orthogonal B-, C-, and D-scan views of the complete data
volume. The data was examined on a beam-by-beam basis, to establish, to start with, that it
was of the correct quality to pass for detailed interpretation and analysis. This involves checking
the MIPS records, the calibration printouts and the Run Logs, and comparing the data images to
ensure that they match. The scan limits and the limits of the processed data volume were also
checked and matched. Then other attributes were checked: scanning surface, beam direction,
material noise levels, presence and location of geometric echoes, presence of electromagnetic
interference or any evidence of other external noise, evidence that the system’s peak detection
capacity has been exceeded, etc..
The data was then passed for analysis, in this case for the detection level and signal-to-noise
ratio for the implanted defects. This is discussed in detail below.
The interpretation of the data was performed by two experienced data interpreters, both Level 2
in MIPS and GUIDE (11.2) and PCN Level 3 in Ultrasonic Inspection of Welds. Both also hold
the BE Inspection Group endorsement for Ultrasonic Inspection of Austenitic Stainless Steel
Components and Welds on BE Plant (11.3) and regularly interpret this type of ultrasonic data.
The MIPS Results files for these inspections have all been supplied to TWI, Cambridge, as the
Lead Partner for the Project, on DVD disks. The Flaw files, GUIDE images and Component
Profile Overlays were also included.
The data has also been archived into the BE Inspection Group system (11.4).

Revision 000
Page 215 of 242
7. Detection Results
7.1 Features of Data (geometric echoes, parent materials, inclusions, attenuation)
The data shows a range of geometric echoes from expected sources: eyebolt holes in top
surface, bottom corners at the ends of the block, etc.. These are largely away from the
inspection volume and do not interfere with the interpretation of the inspection data. They can
help to confirm the correct registration of the inspection system with the datum reference points
on the component.
Assessment of the 0° beams (short range twin, longer range single) shows that the Backwall
Echo on the ferritic side (with the austenitic cladding) is on average 2 dB weaker than on the
austenitic side, but that the austenitic side appears to have a few laminar inclusions spread
around the mid-wall region. These are also seen by a number of the TRL angled beams.
Beams scanning from the ferritic side show evidence of the austenitic cladding on the inside
surface, with a layer of higher amplitude noise:
i) short range, near-surface – beams scanning on inner surface;

ii) long range, far-wall – beams scanning on outer surface – for these beams the noise is
added to by the contribution of the steeper shear wave component sweeping the clad
surface.
The extent of Defect #1 on the austenitic side caused some obscuration of parts of the weld for
beams scanning towards the ferritic side, and it also produced some re-direction effects and
occasionally ‘ghost’ echoes, which were only eliminated by lowering the Pulse Repetition
Frequency (PRF).
7.2 Signal-to-Noise Ratio (SNR) Tables
The detectability of the implanted defects was assessed by measuring the difference between
the peak defect response and the peak grass amplitude local to the defect. These
measurements are presented in Table 2 below, on a defect-by-defect and beam-by-beam basis.
These measurements are inevitably somewhat subjective. Some are straightforward to assess,
some take more manipulation of the images. For these reasons, the measurements are
performed twice, independently, and then the results compared and reconciled.
Unsurprisingly, the highest detection margins are for the defects which beams sweep whilst
passing through little or no weld metal, and with the beam-to-defect angle close to zero. This
can be seen for the inner surface scans hitting defects #4 and #6, which are on the austenitic
side fusion face.
Near-surface defects (#2 for outer surface scans, #3, #5 and #8 for inner surface scans) were
all detected with good margins. Defects #3 and #5 show a strong directional effect due to their
orientation (tilt). Figures 8 and 9 show example data images.

Revision 000
Page 216 of 242
7.3 Penetration of Weld
Defect #7, which is a rough crack embedded in the middle of the weld, is detected by beams in
both directions and from both surfaces, indicating that effective penetration of at least half the
weld width is achievable. Defects #4 and #6 can be detected by beams scanning through the
full width of the weld, but the beams which manage this are mostly from the 1 MHz elliptical
single crystal probes, the detection margins are modest and sometimes detection is via an
indirect response. Figure 10 shows example data images.
Defect #1 (full width slot, 15 mm high) can be detected through the weld, again by the 1 MHz
elliptical single crystal beams (45° and 60°). Figure 11 shows example data images.
7.4 Analysis and Discussion
The usual factors leading to good detection apply: favourable beam-to-defect incidence,
minimising beam path lengths in weld metal and suitable focal lengths and crystal sizes to
maintain effective beam shapes.
Note that, because the access is there, the scans have used the maximum standoff available to
try to reach all parts of the weld with the most effective beam angles. On real plant, often the
‘flat’ on the carbon steel nozzle side is quite short and, for example, the far-wall regions cannot
be scanned with 70° beams from the outer surfaces. In addition, the austenitic stainless steel
safe-end itself may only have a short axial length (typical for US plant) and there may be
another adjacent weld which restricts the access for higher angle beams.
The results for the inspection of these test blocks are consistent with BE experience on
inspecting qualification specimens.
For a weld of this thickness, the typical detection capability claimed for in-service inspection
would be:
a smooth or rough planar defect with dimensions 10 mm throughwall extent by 20 mm long or

larger; surface-breaking or within 10 mm of the surface and within ± 10° tilt or skew of the radial-
circumferential plane.

Revision 000
Page 217 of 242
8. Positioning and Sizing Results
8.1 Additional Sizing Scans
After analysing the search data, a set of sizing scans was compiled to provide the best
positioning, length and throughwall sizing possible. The main principle was to apply beams to
achieve oblique incidence on the defects, in order to record diffracted edge wave signals. This
is generally the most accurate pulse-echo technique for estimating defect throughwall extent.
The downside is that the edge wave signals are usually weak and are not always evident.
Additional 2 MHz 45° TRL probes were used for many of the sizing scans, with an f25 focus
probe used for near-surface scans and an f50 focus probe used for defects in the mid-wall
region. The sizing scans chosen depended on the responses in the search scan data and
generally conformed to the guidelines of applying the beams achieving the shortest range,
ranges closest to the probe focus and shortest path lengths through weld metal.
As noted above, the specific sizing scans implemented a very fine scan pattern of 0.5 x 0.5 mm
and data was collected in a single direction to minimise the effects of backlash in the
manipulator’s mechanism.
Sizing scans carried out:
Block ‘A’ outer surface

ODS21, ODS22, ODS23, ODS24, ODS25, ODS26, ODS27, ODS28
Block ‘A’ inner surface
IDS21, IDS22, IDS23, IDS24, IDS25, IDS26
Block ‘B’ outer surface
ODS25, ODS26, ODS27
Block ‘B’ inner surface
IDS21, IDS22, IDS23, IDS24, IDS28, IDS29, IDS30, IDS31, IDS32, IDS33, IDS34,
IDS35
Defect sizing techniques were matched to the Echodynamic Pattern (EDP) exhibited by the
defects. Most length sizing used the 6 dB amplitude drop method. Most throughwall sizing used
the maximum amplitude method, looking for the maximum in diffracted edge wave signal arcs.
As for the detection assessment, the sizing was performed twice, with the initial results being
verified by a 2nd data interpreter. Very minor differences occurred between the two sets of sizing
results, generally smaller than the differences between one beam and another.
8.2 Defect Positions
For Block ‘A’, the defect start and end positions along the weld appear to be match the drawing
to within approximately 5 mm, mostly 4 - 5 mm low in Y co-ordinate. In Block ‘B’, this difference
increases to match within 8 - 10 mm, possibly due to the material lost in the cut, separating the
original block into two parts.
The smooth defects exhibit fairly clear cut start and ends to their length and the 6 dB drop
technique should be accurate to within a few millimetres.
Some of the defects show very clear top and bottom arcs and very little scatter in the
throughwall extent estimates. Other defects give signals which conflict with the nominal
(drawing) values and are harder to reconcile. Defect #1 interferes with beams scanned from the
austenitic side inner surface.
Revision 000
Page 218 of 242
Defect #8 is smaller than the minimum throughwall extent for which BE Inspection Group would
normally claim sizing capability (for this thickness of weld, this size is usually quoted as 6 mm).
8.3 Best Estimates Defect Sizes
Tables 3 – 9 give the individual measurements for each defect, using the best responses from
the sizing scans.
Table 10 summarises this information into best estimates of the defect positions and extents
and presents the nominal (drawing) values for comparison. The best estimates are based on
judgement of the reliability of particular signals and the level of confidence in results from
different beams.
8.4 Analysis and Discussion
As seen, there are some measurements that are very close to the nominal values and others
where the measurements are significantly different, but, again, this is not untypical of these
kinds of inspections on specimens with implanted defects. There are one or two examples
where additional unexplained responses are seen close to expected defect echoes, and these
need to be included or discounted as part of the sizing process.
Note that the separations of the defects along the weld, i.e. between the end of one and the
start of the next, are very consistent with the drawing, even if the actual positions are a little out.
BE Inspection Group’s experience, borne out by other external and independent exercises, is
that sizing reliability increases when the results of a number of different beams are taken into
account and that diffracted edge wave signals, where they can be detected, provide the most
accurate estimate of a defect’s extremities.
It also needs to be pointed out that, whilst artificial defects are very useful guides to defect
detectability and sizing, real defects sometimes behave differently.

Revision 000
Page 219 of 242
9. Review
The two blocks have been subject to detailed inspection using the MIPS / GUIDE system and a
range of 0° and angled compression wave probes. These inspections have conformed to the
inspection design approach and quality standard typically used for similar inspections within BE.
Scans have been performed in both axial directions, and from both outer and inner surfaces of
the blocks.
The data quality and defect detection capability have been assessed and found to be
acceptable. The data has been repeat interpreted by two experienced data interpreters and
100% verification of results has been applied.
All defects have been detected, although, as expected, the best detection margins are for
defects swept by beams which do not have to traverse significant paths in weld metal. Detection
margins for defects when the beams do have to pass through the weld are lower, but the 1 MHz
elliptical single crystal beams can penetrate the full width of the weld, and still detect defects.
Note that this weld is probably a little wider than typically encountered on real plant, in that the
fusion faces rise 63 mm (85 less 22 = 63) at 15°, whereas the tendency is to try to minimise the
amount of weld metal required. But this does mean that the test welds are a difficult challenge
for ultrasonic inspection.
The sizing performance is seen to be reasonable and is consistent with the tolerances expected
from previous exercises and with the performance claimed for similar inspections on Sizewell B.
The MIPS / GUIDE data has been provided to the Project and the delivery of this report
completes BE’s Action 11 form the 3rd Quarterly Meeting.
10. Conclusions
(1) The two test blocks produced under the DISSIMILAR Project have been inspected using
BE’s MIPS / GUIDE automated ultrasonic inspection system and using methods and
techniques typical of the inspection of this type of joint in nuclear power plant.
(2) The data produced by these inspections provides a ‘conventional’ fingerprint of the welds,
for future comparison with optimised Phased Array inspections.
(3) All the defects have been detected, with varying margins of signal-to-noise ratio.
(4) The defects have been positioned and sized within the expected tolerances for this
thickness of austenitic weld.
(5) The results of all the defect detection assessments and positioning and sizing
measurements have been 100% verified and tabulated in this report.
(6) The delivery of this report completes BE’s Action 11 form the 3rd Quarterly Meeting.

Revision 000
Page 220 of 242
11. References
11.1 EPD/SXB/NDT/1033/00, MIPS/GUIDE Ultrasonic Examination Procedure for the

Automated In-Service Inspection of Steam Generator Channel Head Primary Nozzle to
Safe-End Inlet and Outlet welds, Issue 5, February 2008, BE Inspection Group
11.2 BEGL Inspection Group MS Access Database - MIPS/GUIDE Automated Inspection

Qualifications - G:\Engineering\ETB\Ltasks\Inspection\Inspection Group General
(BPM)/Certification and Registration schemes (Peter Muitt)\British Energy MIPS and
GUIDE Register (SINDT 151)
11.3 BEGL Inspection Group MS Access Database - Austenitic Inspection Qualifications -

G:\Engineering\ETB\Ltasks\Inspection\Inspection Group General (BPM)\Certification and
Registration schemes (Peter Muitt)\British Energy Austenitic Register (SINDT 018)
11.4 E/PROC/ENG/BI/009, Maintenance of Inspection Group Magnetic Media Register,

Revision 001, May 2007

Revision 000
Page 221 of 242
LIST OF FIGURES
FIGURE 1A – Outline of DISSIMILAR Block ‘A’
FIGURE 1B – Outline of DISSIMILAR Block ‘B’
FIGURE 2 – DISSIMILAR Specimen sectioned into Blocks ‘A’ and ‘B’
FIGURE 3 – Etched macrostructure of DISSIMILAR test specimen weld
FIGURE 4 – Scanning Setup - XY Scanning Frame and Immersion Tank
FIGURE 5 – Scanning Co-ordinate System
FIGURE 6 – Fine-grain forged austenitic stainless steel calibration blocks
FIGURE 7 – Inspection Volume
FIGURE 8 – Example GUIDE Image – Defect #3 – 1 MHz 45° TRL1 f25 – ID005
FIGURE 9 – Example GUIDE Image – Defect #5 – 1.5 MHz 70° TRL1.5 f35 – ID013
FIGURE 10 – Example GUIDE Image – Defect #4 – 1 MHz 60° L1 – OD009
FIGURE 12 – Typical PWR RPV Nozzle Safe-End Weld Design

Revision 000
Page 222 of 242
FIGURE 1A – Outline of DISSIMILAR Block ‘A’

Revision 000
Page 223 of 242
FIGURE 1B – Outline of DISSIMILAR Block ‘B’

Revision 000
Page 224 of 242
FIGURE 2 – DISSIMILAR Specimen sectioned into Blocks ‘A’ and ‘B’
FIGURE 3 – Etched macrostructure of DISSIMILAR test specimen weld
Buttering
Inconel
Carbon weld metal Stainless
steel steel
Cladding

Revision 000
Page 225 of 242
FIGURE 4 – Scanning Setup - XY Scanning Frame and Immersion Tank
FIGURE 5 – Scanning Co-ordinate System

Z=0
Cross-Section
Carbon Steel Stainless Steel

('Nozzle') ('Safe-End') +Z
-X +X
Y=0
+Y

Plan
('Nozzle') ('Safe-End')
X = 0 Datum at
Weld Centreline

Revision 000
Page 226 of 242
FIGURE 6 – Fine-grain forged austenitic stainless steel calibration blocks

Revision 000
Page 227 of 242
FIGURE 7 – Inspection Volume
15

('Nozzle') ('Safe-End')

Revision 000
Page 228 of 242
FIGURE 8 – Example GUIDE Image – Defect #3 – 1 MHz 45° TRL1 f25 – ID005
High Cladding Noise
FIGURE 9 – Example GUIDE Image – Defect #5 – 1.5 MHz 70° TRL1.5 f35 – ID013

Revision 000
Page 229 of 242
Defect #4
Defect #1

Revision 000
Page 230 of 242
FIGURE 12 – Typical PWR RPV Nozzle Safe-End Weld Design
Forged Ferritic
Forged Austenitic Pipe Forged Austenitic Safe-End Outlet Nozzle
(SA376, 304LN) (SA182, 316LN) (SA508, Cl.3)
Outlet
(Hot Leg)
Buttering Cladding
SE
1
PV
W
C
Statically Cast Forged Ferritic

Austenitic 22° Elbow Forged Austenitic Safe-End Inlet Nozzle
(SA351, CF-8) (SA182, 316LN) (SA508, Cl.3)
Inlet
(Cold Leg)
Buttering Cladding
SE
14
PV
W
C
Note: 1) short length of safe-end and consequent proximity of pipework welds

2) limited flat on nozzle side outer surface, before blend radius

Revision 000
Page 231 of 242
LIST OF TABLES
TABLE 1 – List of Probes / Beams applied to DISSIMILAR Test Specimen Weld Blocks
TABLE 2 – Detection and Signal-to-Noise Ratios for Defects in Blocks ‘A’ and ‘B’
TABLE 3 – Positioning and Sizing Measurements – Defect #2 Block ‘A’
TABLE 6 – Positioning and Sizing Measurements – Defect #5 Block ‘B’
TABLE 10 – Best Estimate of Defect Positions and Extents

Revision 000
Page 232 of 242
TABLE 1 – List of Probes / Beams applied to DISSIMILAR Test Specimen Weld Blocks
Single/ Casing/mm
Number Make Angle/° Freq/MHz Focus/mm Crystal(s)/mm Scans
Twin (LxWxH)
Applus- 2(½ Ø10)
1 0 2.5 Twin 12 40x40x35 Search
RTD Composite
Applus- Ø25
2 0 1 Single -- 40x40x35 Search
RTD Composite
Applus- 2(15x25)
3 70 1.5 Twin 35 40x40x35 Search
RTD Composite
Phoenix Ell(25x43)
4 70 2.5 Single -- 60x40x40 Search
ISL PZT ¼λ
Applus- Ell(25x43)
RTD Composite
Applus- Ell(25x43)
RTD Composite
2(15x25)
7 RTD 60 2 Twin 35 40x40x35 Search
PZT
2(15x25)
8 RTD 60 2 Twin 65 40x40x35 Search
PZT
Applus- Ell(25x34)
RTD Composite
Applus- 2(15x25)
10 45 1 Twin 25 40x40x35 Search
RTD Composite
2(15x25)
11 RTD 45 2 Twin 50 40x40x30 Sizing
PZT ¼λ
2(10x18)
12 RTD 45 2 Twin 25 40x40x35 Sizing
PZT
All compression wave probes

Revision 000
Page 233 of 242
TABLE 2 – Detection and Signal-to-Noise Ratios for Defects in Blocks ‘A’ and ‘B’
DEFECTS
BEAM SURFACE DIRECTION BOTH BLOCK A BLOCK B
1 2 3 4 5 6 7 8
1.5 MHz OD014 OD014

OD Aust→Ferr NB 30 NB 6 NB NB NB NB
70° f35
1.5 MHz OD013 OD013 OD013

OD Ferr→Aust NB 12 NB 10 NB 6 NB NB
70° f35
1.5 MHz ID014 ID014 ID014 ID014 ID014

ID Aust→Ferr 22+ NB ND 16 6 20 ND 9
70° f35
1.5 MHz ID013 ID013 ID013 ID013 ID013 ID013

ID Ferr→Aust 20+ NB 18 6 20+ ND 12 17
70° f35
2 MHz OD016
OD Aust→Ferr NB 14 NB ND NB ND ND NB
60° f35
2 MHz OD015
OD Ferr→Aust NB 6 NB ND NB ND ND NB
60° f35
2 MHz ID016 ID016 ID016

ID Aust→Ferr 12 NB 8 NB ND ND ND 19
60° f35
2 MHz ID015 ID015 ID015 ID015 ID015

ID Ferr→Aust 10 NB 20+ NB 19+ ND 10+ 16+
60° f35
1 MHz OD006
OD Aust→Ferr NB 14+ NB ND NB ND ND NB
45° f25
1 MHz OD005
OD Ferr→Aust NB 6 NB ND NB ND ND NB
45° f25
1 MHz ID006 ID006 ID006 ID006 ID006

ID Aust→Ferr 12 NB 6 6 9 NB ND 15
45° f25
1 MHz ID005 ID005 ID005 ID005 ID005

ID Ferr→Aust 12 NB 24+ NB 12 NB 18 14
45° f25
1 MHz OD008 OD008 OD008 OD008 OD008 OD008

OD Aust→Ferr 24 NB 15 6 8+ 10 10 ND
45° Single
1 MHz OD007 OD007 OD007 OD007 OD007

OD Ferr→Aust 9 NB 8 9 6 8 ND ND
45° Single
1 MHz
ID Aust→Ferr NB ND NB ND NB ND ND NB
45° Single
1 MHz ID007 ID007

ID Ferr→Aust NB 6* NB ND NB ND 8 NB
45° Single
2 MHz OD018 OD018 OD018 OD018 OD018
OD Aust→Ferr NB 22 15 ND 9+ 6 12 NB
60° f65

Revision 000
Page 234 of 242
DEFECTS
BEAM SURFACE DIRECTION BOTH BLOCK A BLOCK B
1 2 3 4 5 6 7 8
2 MHz OD017 OD017

OD Ferr→Aust NB 10 ND ND ND 6+ ND NB
60° f65
2 MHz ID018 ID018 ID018
ID Aust→Ferr NB ND ND 24 ND 24 8 NB
60° f65
2 MHz ID017 ID017 ID017 ID017
ID Ferr→Aust NB 12 15 ND 20+ ND 22 NB
60° f65
OD Aust→Ferr 20 NB 9 12 * 8+ 8+ * 8+ ND
60° Single
1 MHz OD009 OD009 OD009 OD009
OD Ferr→Aust 10 NB 10* 10 ND 8+ ND ND
60° Single
1 MHz ID010 ID010
ID Aust→Ferr NB ND NB 20+ NB 21 ND NB
60° Single
1 MHz ID009 ID009
ID Ferr→Aust NB 12 * NB ND NB ND 12 * NB
60° Single
OD Aust→Ferr 16 NB 10 10 12 * 12 * 14 ND
65° Single
1 MHz OD011 OD011 OD011 OD011
OD Ferr→Aust 14 NB 8 10 * ND 8 ND ND
65° Single
1 MHz ID012 ID012 ID012
ID Aust→Ferr NB ND NB 28 NB 23 20 * NB
65° Single
1 MHz ID011 ID011
ID Ferr→Aust NB 10 * NB ND NB ND 15 NB
65° Single
2.5 MHz OD020 OD020 OD020 OD020 OD020
OD Aust→Ferr NB NB ND 14 * 15 7* 12 9
70° Single
2.5 MHz OD019 OD019
OD Ferr→Aust NB 10 ND ND NB 6* ND ND
70° Single
2.5 MHz ID020 ID020 ID020
ID Aust→Ferr NB ND NB 36+ NB 35 * 6* NB
70° Single
2.5 MHz ID019 ID019 ID019
ID Ferr→Aust NB 8 NB ND 18 ND 15 NB
70° Single
Key: NB – Beam does not sweep defect depth, or beam not gated for this depth / range
ND – Defect not detected
* – higher amplitude signal also present from associated mechanism, e.g. self-tandem

Revision 000
Page 235 of 242
BLOCK A - DEFECT DEFECT

DEFECT 2 THROUGHWALL EXTENT (mm) LENGTH (mm)
Scan / Surface Top Bottom Height Ligament Start End Length
IDS21
10.3 mm to
2 MHz 45° f50 10.3 19.2 8.9 83.6 103.1 19.5
OD
ID - Ferr→Aust
Maximum Amplitude sizing used 6 dB sizing used
IDS22
3.9 mm to
2 MHz 60° f65 3.9 14.9 11 84 104 20
OD
ID - Ferr→Aust
6 dB sizing used, no tip arc evident.
Sizing errors probably due to proximity of
6 dB sizing used
backwall and beam passing through
cladding.
ODS21
8.6 mm to
2 MHz 45° f25 8.6 20.1 11.5 79.5 102.5 23
OD
OD - Ferr→Aust
ODS22
8.9 mm to
2 MHz 45° f25 8.9 22.4 13.5 81 106.5 25.5
OD
OD - Aust→Ferr
ODS23
11 mm to
2 MHz 60° f35 11 19 8 81 102.5 21.5
OD
OD - Ferr→Aust
ODS24
8 mm to
2 MHz 60° f35 8 20 12 80.5 106 25.5
OD
OD - Aust→Ferr

Revision 000
Page 236 of 242

DEFECT 3 THROUGHWALL EXTENT(mm) LENGTH (mm)
IDS25
66.9 mm
2 MHz 45° f25 66.9 75.9 9 130.5 156 25.5
to OD
ID - Ferr→Aust
IDS26
69.8 mm
2 MHz 60° f35 69.8 80.3 10.5 131 154.5 23.5
to OD
ID - Ferr→Aust
ODS27
64.5 mm
2 MHz 45° f50 64.5 77 12.5 130 153 23
to OD
OD - Aust→Ferr
6 dB sizing used 6 dB sizing used
ODS28
60.1 mm
1 MHz 45° Single 60.1 77.6 17.5 129.5 156.5 27
to OD
OD - Aust→Ferr

Revision 000
Page 237 of 242

DEFECT 4 THROUGHWALL (mm) LENGTH (mm)
IDS23
33.6 mm
2 MHz 45° f50 33.6 50.6 17 184.5 203 18.5
to OD
ID - Aust→Ferr
Maximum Amplitude sizing used. Beam
partially obscured by Defect 1, resulting in 6 dB sizing used
probable mis-positioning of indication
IDS24
20.5 mm
2 MHz 60° f65 20.5 38 19.5 182 198.5 16.5
to OD
ID - Aust→Ferr
Throughwall extent greater than expected
theoretical defect. Unable to define 6 dB sizing used
alternate signal mechanism for indication.
ODS25
37.2 mm
2 MHz 45° f50 37.2 48.2 11 180 198 18
to OD
OD - Ferr→Aust
ODS26
39 mm to
2 MHz 45° f50 39 52.6 13.6 NA NA NA
OD
OD - Aust→Ferr
Poor signal response, not
Maximum Amplitude sizing used
able to assess length

Revision 000
Page 238 of 242
BLOCK B - DEFECT DEFECT

IDS21
58.8 mm
1.5 MHz 70° f35 58.8 75.8 17 68.5 91.5 23
to OD
ID - Ferr→Aust
IDS22
61.6 mm
2 MHz 60° f35 61.6 74.1 12.5 67.5 91.5 24
to OD
ID - Ferr→Aust
IDS23
62 mm to
2 MHz 45° f25 62 74 12 70.5 92 21.5
OD
ID - Ferr→Aust

IDS24
41.8 mm
2 MHz 45° f50 41.8 54.8 13 115 139.5 24.5
to OD
ID - Aust→Ferr
Maximum Amplitude sizing used. Beam
obscured by Defect 1 resulting in beam
6 dB sizing used
edge detection only, causing mis-
positioning of indications
ODS25
39.8 mm
2 MHz 45° f50 39.8 50.8 11 122.5 150.5 28
to OD
OD - Aust→Ferr

Revision 000
Page 239 of 242

ODS26
34.3 mm
2 MHz 45° f50 34.3 NA NA 178 189.5 11.5
to OD
OD - Ferr→Aust
Single arc, insufficient S/N to use 6dB
6 dB sizing used
sizing
ODS27
42.5 mm
2 MHz 45° f50 42.5 47 4.5 175 190.5 15.5
to OD
OD - Aust→Ferr
IDS28
40.7 mm
2 MHz 45° f50 40.7 51.7 11 164 191 27
to OD
ID - Ferr→Aust
IDS29
38.2 mm
2 MHz 45° f50 38.2 48.2 10 169 192 23
to OD
ID - Aust→Ferr

Revision 000
Page 240 of 242

IDS30
80 mm to
1.5 MHz 70° f35 80 NA 5 219.5 238.5 19
OD
ID - Ferr→Aust
Single arc (truncated) Maximum
6 dB sizing used
Amplitude sizing used
IDS31
77.5 mm
1.5 MHz 70° f35 77.5 NA 7.5 221.5 242 20.5
to OD
ID - Aust→Ferr
6 dB sizing used
IDS32
82.2 mm
2 MHz 60° f35 82.2 NA 2.8 218.8 240 21.5
to OD
ID - Ferr→Aust
6 dB sizing used
IDS33
79.1 mm
2 MHz 60° f35 79.1 NA 5.9 220.5 242.5 22.5
to OD
ID - Aust→Ferr
6 dB sizing used
IDS34
79.7 mm
2 MHz 45° f25 79.7 NA 5.3 217.5 239 21.5
to OD
ID - Ferr→Aust
IDS35
2 MHz 45° f25 NA NA NA NA NA NA` NA
ID - Aust→Ferr
Defect exhibits poor S/N unable to give accurate dimensions

Revision 000
Page 241 of 242
TABLE 10 – Best Estimate of Defect Positions and Extents
DEFECT NOMINAL / DEFECT LENGTH / mm DEFECT THROUGHWALL EXTENT / mm

BLOCK
NUMBER MEASURED Start End Length Top Bottom Height Ligament
NOMINAL 85 110 25 10 18 8 10
2
MEASURED 81 106 25 9 20 11 9
NOMINAL 135 160 25 67 75 8 67

A 3
MEASURED 129 155 26 67 78 11 67
NOMINAL 185 210 25 38 48 10 38

4
MEASURED 180 201 21 34* 49* 15* 34
NOMINAL 77 102 25 65 75 10 65
5
MEASURED 69 92 23 62 74 12 62
NOMINAL 126 151 25 40 45 5 40

6
MEASURED 116 140 24 41* 51* 10* 41
B
NOMINAL 176 201 25 37 49 12 37
7
MEASURED 167 194 27 41 52 11 41
NOMINAL 226 251 25 82 85 3 82

8
MEASURED 219 240 21 80 85 5 80
* - Defects #4, #6 partly obscured by Defect #1 from austenitic side and some anomalous
signals noted which were difficult to reconcile – see Tables 5, 7

Revision 000
Page 242 of 242
DISTRIBUTION / NOTIFICATION LIST
Capucine Carpentier Project Manager - DISSIMILAR TWI, Cambridge
E/TSK/GEN/7228 BE Inspection Group Task File Barnwood

ISO 9001 2015 Training

Uploaded by

Copyright:

Available Formats

ISO 9001 2015 Training

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

ISO 9001 2015 Training

Uploaded by

Copyright:

Available Formats

Technology Strategy Board

Guidelines for generating array ultrasonic procedures for the

MATERIALS FOR ENERGY

Phased Array Ultrasonic Inspection of Dissimilar Metal

TSB Project No: TP11/MFE/6/I/AA058J

Date: 25 January 2011 Ref: TP AA058J-FR

Circulation: Lead organisation: TWI Ltd

Project Monitoring Officer Alan Groves (dstl)

Guidelines for generating array ultrasonic procedures for

2. Ultrasonic technique development should be performed on a representative sample

3. The velocity of ultrasound is dependent on propagation direction within an austenitic

The adapted delay law (ADL) technique

Personnel and training

performance of the personnel must be assessed to ensure they meet minimum

Validation and qualification

Key parameters for technique design

2.1 JOINTS WITH AUSTENITIC DEPOSITS 17

3.1 INTRODUCTORY COMMENTS 25

4.1 INTRODUCTORY COMMENTS 51

6.1 INTRODUCTORY COMMENTS 86

7.1 INTRODUCTORY COMMENTS 95

10 FUTURE DIRECTIONS 159

10.1 ARRAY PROBE CAPABILITIES 159

Appendix A: Ultrasonic inspection of an austenitic weld - a case study

The DISSIMILAR project (www.dissimilarweld.co.uk) was initiated in July of 2008 and

The aim of the project was to:

The industrial case

Ultrasonic phased array technology

Phased array ultrasonic technology is a sophistication of the conventional ultrasonic

Contributing authors and their affiliations

Bartlett , Alan Applied Inspection

It is noted as a key recommendation that, ideally, a holistic design of an austenitic joint -

The particular welded component considered in the DISSIMILAR project is described in

2.1 Joints with austenitic deposits

2.2 Joint configuration of the DISSIMILAR project

Simulating typical weld flaws

Figure 1 Weld preparation of the DISSIMILAR specimen.

Figure 2 Details of the procedure for cladding the 50D plate.

Process GTAW and SMAW.

Figure 4 Details of the procedure for filling the weld.

Figure 14 shows the as-received specimen post-fabrication.

Figure 6 Flaw 1, EDM slot in the stainless steel parent material.

Figure 9 Flaw 4, smooth LOF on the weld/stainless steel fusion face.

Figure 11 Flaw 6, smooth LOF on the weld/stainless steel fusion face.

Figure 13 Flaw 8, smooth lack of root fusion.

Figure 14 Image of the as-received DISSIMILAR test specimen.

3.1 Introductory comments

Ferritic nozzle - 5892m/s

3.2 EBSD approach to quantifying the weld

Figure 18 Position of the three samples, referred to as Sample 1, 2 and 3.

3.3 Optimisation of EBSD scanning parameters

3.3.1 The minimum scatterer criterion

Figure 22 Through transmission experimental setup to measure the attenuation.

3.3.2 Empirical study of scanning resolutions

3.4 Processing the EBSD data

3.5 Analysis of the microstructure and texture of the austenitic weld

weld microstructure varies parallel to the welding direction, ie in a direction perpendicular to

3.6 Lengthwise uniformity

3.6.1 Analysis of lengthwise uniformity using EBSD